Journal of Cleaner Production 418 (2023) 138146

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

The synergy effect through combination of the digital economy and

transition to renewable energy on green economic growth: Empirical study
of 18 Latin American and caribbean countries
Young Kyu Hwang
Autonomous University of Madrid, Spain

A R T I C L E I N F O A B S T R A C T

Handling Editor: Giovanni Baiocchi Tremendous effects on the global economy in terms of economic, social and environmental costs remind us of the
catastrophic consequences of climate change and global warming caused by CO2 emissions. Therefore, accel­
Keywords: erating decarbonization of the global energy system should be put in place to curb large amount of CO2 emission
Synergistic effect from hydrocarbon energy sources on which the current global value chain of production heavily relies.This study
Renewable energy transition
focuses on analyzing the effects of renewable energy transition, the digital economy, and the synergy between
Digital economy
them on green economic growth in 18 Latin American countries. To capture the multidimensionality of these
Green economic
Growth transitions, the Renewable Energy Transition Index (RETI) and the Digital Economy Index (DEI) are developed
Latin America using the Principal Component Factor (PCF) technique. The FixedEffect Panel Threshold Regression (FEPTR)
substantiates that renewable energy transition has a significant threshold effect on economic growth and envi­
ronmental sustainability depending on the level of income and carbon emissions. On the other hand, the Method
of Moments Quantile Regression (MMQR) shows that both renewable energy transition and the digital economy
have a significant positive impact on economic growth in all quantile groups. For the environmental sustain­
ability, only renewable energy transition is found to have a positive impact in all quantile groups. From the
synergistic effect perspectives, the CO2 emissions reduction is observed in both the low and middle quantile
groups, but the economic growth promotion is only observed in a low quantile group.

1. Introduction carbon budget with a 67% probability to constraint a rise in global

According to the Intergovernmental Panel on Climate Change’s Sixth
Assessment Report (IPCC, 2022), the global average temperature has
been rising approximately 1.1◦ C compared with that of the late 1800s.
Specifically, the last decade from 2010 to 2019, the average temperature
on earth was at a record high without precedents in comparison with the
previous decades. The main factors of this global warming can be found
in GHG emissions, especially in CO2 emissions caused by anthropogenic
activities and the burning of fossil fuels. Indeed, the average annual GHG
emissions during 2010–2019, coinciding with the warmest temperature
registered during this decade, were higher than ever even though the
average annual growth rate of GHG emissions has been slowed down
compared to that of 2000–2009 (1.3% against 2.1%) (IPCC, 2022). It is
worth noting that the historical cumulative net CO2 emissions from 1850
to 2019 account for almost four fifths of the total carbon budget with
50% probability to restrain the increase in global average temperature
to well below 1.5◦ C above preindustrial levels and two thirds of the total
average temperature to well below 2◦ C above pre-industrial levels
(IPCC, 2022). Since the remaining carbon budget to meet the Paris
Agreement Goals is quite limited in order to avoid catastrophic and
irreversible consequences of global warming and climate change, poli­
cymakers and researchers agree on taking urgent measures to deal with
this issue (World Bank, 2022). As viable alternatives to the issue, they
are managing, a paradigm shifts from fossil fuels towards clean and
renewable energy sources, known as renewable energy transition, is
considered as one of the most prominent and promising option to
dealing with global warming and CO2 emissions reduction (Smil, 2020).
This is because the energy supply is the main sector responsible for
global GHG emissions which account for approximately 34% of total
GHG emissions worldwide in 2019, followed by industry (24%), agri­
culture, forestry, and other land use (22%), transport (15%), and resi­
dential and commercial sectors (6%) (IPCC, 2022), thus the importance
of deep decarbonization in energy sector cannot be overemphasized.
The key role that renewable energy sources and clean energy

E-mail address: young.hwang@estudiante.uam.es.

https://doi.org/10.1016/j.jclepro.2023.138146
Received 25 February 2023; Received in revised form 12 July 2023; Accepted 16 July 2023
Available online 19 July 2023
0959-6526/© 2023 The Author. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
Y.K. Hwang
technologies can play in achieving global net-zero goals and moving
towards a low-carbon economy is also well summarized in the Kaya
identity (Kaya, 1990). According to this theory, carbon emissions are
determined by three major factors: the carbon intensity of energy
sources (C/E), energy intensity (E/Y), and economic output or growth
(Y).1 To mitigate the negative impact of energy use and economic ac­
tivities on environmental quality, significant changes should be made in
terms of the energy mix and sectoral shift, namely moving towards
low-carbon energy sources such as renewables and low-energy intensive
sectors while maintaining economic growth (Hübner, 2018).
In the new decarbonization scenario, the renewable energy sources
are expected to have a prominent role in the coming decades, the digital
economy has recently been drawing much attentions from many gov­
ernments, academics and policymakers who recognized it not only as an
important tool for fulfilling a rapid and prompt low-carbon economy
transition required to avoid catastrophic consequences on the planet and
ensure environmental sustainability, but also as a key enabler to
improving countries’ competitiveness and productivity in the global
market for long-term economic growth (Nwaiwu, 2021). This is because
the digital economy has been expecting to greatly facilitate the inte­
gration of variable renewables in energy mix by relieving their inherent
intermittency and variability through accurate predictions of renewable
energy production (Lin and Huang, 2023) and efficient management of
energy and resource use via application of smart technologies such as
Artificial Intelligence (AI), big data, cloud computing, machine learning
and block chain in energy- and emission-intensive sectors (mainly in­
dustry, manufacturing and construction, transport, electricity and
heating sector) (IEA, 2017). It can also promote green innovations,
which allow to save energy enormously and input costs in the produc­
tion process, reduce energy consumption and GHG emissions, and
improve productivity of workers (Asif, 2020). The potential impact of
digitalization on decoupling economic growth and energy use from
environmental degradation in the process of transition to renewable
energy can be closely linked to the Environmental Kuznets Curve (EKC)
hypothesis (Grossman and Krueger, 1995). According to this theory, a
country’s emission level at the initial stage of industrialization increases
rapidly due to growing economic activities accompanied by the use of
polluting energy sources and low energy efficiency (scale effect). How­
ever, after reaching to a certain level of income and technology level, the
environmental quality of a country begin to improve due to the gradual
adoption of more efficient and environmentally friendly technologies
(technical effect) in the production process and a structural trans­
formation from high-energy intensive sectors such as manufacturing
towards a more knowledge based service sectors (composition effect).
Since the adoption of digital technologies allows countries to greatly
improve their total factor productivity and achieve rapid economic
growth, it can trigger green innovation efforts and facilitate the adoption
of advanced clean energy technologies in the production process. This
can ultimately lead to decoupling of economic growth and energy use
from environmental degradation as predicted by the EKC hypothesis.
Even though many benefits are expected from the synergy between
renewable energy transition and the digital economy on green economic
growth in developing countries, most of the previous studies were
centered around developed or large emerging countries such as China
and India. In this context, this study examines how the combination of
renewable energy transition and the digital economy can contribute to a
substantial economic growth and reductions in carbon emissions in
Latin American and Caribbean (LAC) countries and analyzes the asym­
metric and heterogenous impact of transition to renewable energy in
association with income and carbon emissions, given that LAC countries
are highly different from energy mix, production structure, and natural
resources endowment. Thanks to digitalization, accelerating the
1
C–– C/E x E/Y x Y, where C, E and Y denote carbon emissions, energy use
and economic output resepectively.
2
Journal of Cleaner Production 418 (2023) 138146
integration of renewable energy sources into the energy mix can relieve
the energy trilemma facing the LAC region; energy affordability, energy
security, and energy sustainability (OLADE, 2020). This is due to the fact
that application of digital technologies in renewable energy power grids
(smart grids) facilitate modern, reliable and affordable energy access by
the people living in remote areas, and improve their living conditions
(IEA, 2017). Furthermore, it can trigger green innovation efforts and
bring about significant improvements in energy productivity and effi­
ciency, cost reductions and energy savings (Luo et al., 2022), which will
bring great repercussions upon production capabilities and energy sus­
tainability (Fankhauser and Jotzo, 2018). More on that, the digitaliza­
tion can provide more flexibility, reliability, and resilience to renewable
energy systems through sector coupling, which will facilitate sharing
information and data among different sectors, and enhancing energy
security (Ren et al., 2021). Based on the critical analysis of observations
in the previous studies and the research so far, the main research
problems of this study are figured out as follows.
• Does the synergistic effect between the transition to renewable en­
ergy and the digital economy for green economic growth exist in the
LAC region? In other words, does the synergistic effect between them
help to successfully decouple economic growth from environmental
degradation?
• Are there any significant non-linear and threshold effect of the
transition to renewable energy on economic growth and environ­
mental quality in association with a country’s income and emission
levels in the LAC region?
• Does the impact of the transition to renewable energy, the digital
economy, and the interaction between them on economic growth
and environmental quality show significant heterogeneity across
different quantile groups in the LAC region?
The salient results from our responses to the research problems
above are followed by: First, the Renewable Energy Transition Index
(RETI) and the Digital Economy Index (DEI) specific to the LAC region
are developed to evaluate the impact on green economic growth by
encompassing multidimensional features and complex reality accom­
panied by renewable energy transition and the digital economy. The
previous studies rely on the one-dimensional indicator (renewable en­
ergy consumption or renewable energy generation as a proxy for
renewable energy transition and Internet penetration rate or mobile
subscriptions as a proxy for digital economy) had their limit in
describing complex reality and multidimensional features from which
renewable energy transition and the digital economy entail. Second, the
interaction term between renewable energy transition and the digital
economy is incorporated into the regression models to estimate the
potential synergistic effect between the digital economy and renewable
energy transition on promoting economic growth and enhancing envi­
ronmental quality in the LAC region. Thanks to ever evolving digital
technologies and their popular use in our daily lives, it is right time for
this study to get initiated, given that the governments in the LAC region
are challenging viable approaches to successful implementation of
decarbonization strategy by leveraging potentials of the digital econ­
omy. Third, given that the LAC region consisting of many countries with
big difference in their income levels, emission intensity, energy system,
resource endowments, and structural composition, the impact of tran­
sition to renewable energy and the digital economy on economic growth
and environmental quality is expected to be quite different. Therefore,
the FEPTR and the MMQR econometric techniques are used to analyze
asymmetry and non-linearity features of their impact across Latin
American and Caribbean economies. The analysis would provide gov­
ernments and policymakers of each country alike with invaluable in­
formation concerning about a specific challenges and opportunities for
low-carbon economy transition.
The remaining parts of this paper are organized as follows: Section 2
discusses the potential opportunities and challenges that the LAC region
Y.K. Hwang
would come across during the process of transition to renewable energy
and the digitalization of their economies. Section 3 presents a literature
review. Section 4 describes the methodology and the dataset used in this
study. Section 5 discusses empirical results and the findings in this study.
Section 6 provides conclusions and policy implications.
2. Background
2.1. Renewable energy transition in the LAC
In the LAC region, energy transition is not just simply a shift from
high-carbon intensive energy sources such as fossil fuels (coal, oil, and
natural gas) to low-carbon intensive ones such as renewable energies
(solar, wind, geothermal, tidal, green hydrogen), but also implies a
profound and fundamental structural transformation in social, eco­
nomic, and environmental dimensions which lead to a radical change in
the way people produce and consume energy (Guimarães, 2020). This
paradigm shifts of renewable energy transition present both opportu­
nities and challenges for the LAC region. The main reasons that
renewable energy transition might benefit and offer a great opportunity
for the LAC countries to fulfill sustainable development goals (economic,
social, and environmental development) are the following. First,
renewable energy transition can contribute to achieving the goal of
universalization of modern energy access (electricity) in the region,
especially in remote areas where network infrastructure does not reach
to providing reliable electricity services through decentralized and
off-gird renewable energy systems such as distributed solar photovol­
taic, wind turbines (Vanegas Cantarero, 2020). Although in the LAC
region, 95% of population has already access to electricity (about 18.1
million people were still lacking electricity in 2018), approximately a
third of those electricity connections are illegal and electricity theft
occurs frequently (Guimarães, 2020). In this aspect, the decentralized
and off-grid renewable energy system might be a good alternative to
providing reliable electricity services to those people who live in remote
and marginalized areas where the electricity cannot reach due to the
lack of electricity network and their geographical location (Vanegas
Cantarero, 2020). Also, the decentralized and off-grid renewable energy
system can encourage consumers to actively participate in energy gen­
eration and consumption process (consumers become prosumers),
incentivize people to manage their own energy bills more wisely and
reduce energy thefts which contributes to more revenue streams and
local communities’ socioeconomic development (Asif, 2020). Second,
Renewable energy transition can contribute to reducing energy de­
pendency of countries and thus enhancing energy security, especially for
net importer of crude oil and oil products such as many Caribbean and
Central American countries by reducing their vulnerability to high- and
volatile oil prices in the global spot market, thus contributing to lower
trade and fiscal deficits (Fattouh et al., 2019). Some net oil export
countries such as Brazil, Ecuador, and Venezuela can also benefit from
renewable energy transition as the increasing participation of renew­
ables in energy system might contribute to diversification of energy mix
and thus improving their energy security (Henderson and Sen, 2021). In
sum, the renewable energy transition can bring about macroeconomic
stability via lowering high vulnerability to commodity prices in the
global market and provide more resilience to energy system through
diversification of energy mix in the LAC region. Third, renewable energy
transition can contribute to sustainability improvement by substituting
fossil fuel energy sources with renewable ones, thus raising the share of
renewable energy in energy mix which results in reducing CO2 emissions
and environmental degradation (Hampl, 2022). Fourth, renewable en­
ergy transition might contribute to reducing energy poverty and
inequality in the LAC region through offering affordable energy services
to vulnerable population as marginal cost of producing electricity from
renewables are close to zero (Urban, 2014). Thanks to this cheaper
electricity price, the economic burden of poor households lessens, and
consequently their purchasing power also increases (Vanegas Cantarero,
3
Journal of Cleaner Production 418 (2023) 138146
2020). Lastly, due to the high- and untapped renewable energy potential
of the LAC region, it might attract huge investments from multinational
companies and private investors. Due to the Keynesian multiplier effect
of these investments, renewable energy transition might have a positive
impact on countries’ economic performance as well as on environmental
quality promoting green growth in the region (Hafner and Luciani,
2022). On the flip side, renewable energy transition also presents some
challenges for the LAC region for the following reasons. First, the large
part of government budget of big oil producing and exporting countries
in the region such as Brazil, Mexico, Argentina, Colombia, and
Venezuela are heavily dependent on revenues stemming from oil, and
they might incur huge fiscal deficit due to a drastic loss of revenues
during the process of renewable energy transition (Fattouh et al., 2019).
Second, high debt burden in many LAC countries constraints severely
their fiscal capacity limiting public investments in renewable energy
transition, therefore, many investors from private sectors may be
reluctant to finance large-scale renewable energy deployment in the LAC
region as they perceive high risk and uncertainty regarding their returns
on investments (CEPAL, 2022a). Third, fossil fuel subsidies are already
common and widespread practice in many LAC countries which makes
the price of fossil fuels artificially lower compared to other such as
non-conventional renewable energy sources (solar, wind) at their initial
stage of development. Consequently, non-conventional renewable en­
ergy sources cannot compete with fossil fuels on a level playing field in
national energy market hampering a rapid low-carbon energy transition
necessary to achieve net-zero emission goals in the region (Urban,
2014). Fourth, large endowment of oil and gas reserves (after Middle
East, the LAC region has the second largest reserves of oil and gas in the
world) along with existing production system strongly based on hy­
drocarbons such as fossil-fueled thermal power plants with long lifecycle
might lead to lock-in carbon-intensive energy sources and disincentivize
efforts towards renewable energy transition in the short- and
medium-term in the LAC countries (Hampl, 2022), thus delaying
significantly decarbonization of their economies. Furthermore, the lack
of affordable energy storage technologies so far (lithium-ion batteries
and hydrogen energy are still not cost competitive enough for large-scale
application) indispensable to deal with intermittent and variable fea­
tures of renewable energies makes it even harder to implement a rapid
decarbonization in the LAC region (Smil, 2020). Lastly, weak institu­
tional systems characterized by high corruption level, low level of de­
mocracy, lack of stringent environmental laws and regulations, high
influence of large oil companies on political decision-making process in
favor of their benefits, and low awareness of urgency of environmental
degradation and climate change issues among policymakers and pop­
ulations makes it harder to phase out rapidly fossil fuels in their energy
systems and prevents LAC countries from moving to renewable energy
transition (Vanegas Cantarero, 2020).
As for the energy supply systems in the LAC region, this region has
one of the cleanest electricity mixes in the world in large part due to the
high share of hydropower (Grottera, 2022). Most energy sources used in
electricity generation come from non-polluting renewable energy sour­
ces such as hydro, solar, wind, and geothermal (OLADE, 2020).
Concretely, in 2019, 58.5% of the total power generation in the LAC
region was derived from renewables within which hydro accounts for
45.2% followed by wind (6%), renewable thermal energy (5.1%), Solar
(1.5%), and Geothermal (0.7%) while the share of renewables in power
generation was 26.8% in worldwide (OLADE, 2020). Furthermore, the
total installed capacity of renewables in electricity sector accounted for
58.9% in 2018 with non-conventional renewable energy sources such as
wind and solar experiencing remarkable growth from 2010 to 2018
(from 0.5 to 5.9% in case of wind energy and while solar energy reached
2.1% of participation in 2018) (Messina, 2020). From the demand-side
perspective, the region is also characterized by a high participation of
renewables in terms of total energy consumption (although the figure is
much higher from the supply side than from the demand side). For
instance, in the LAC region, 29% of total energy consumption came from
Y.K. Hwang
renewables in 2018 while in worldwide, it only represented by 16%
(Pablo Romero et al., 2022). The comparatively high participation of
renewable energy in both supply and demand side along with a low
energy consumption per capita in relative terms compared with that of
other regions in the world makes the LAC region a relatively lower
contributor to global net CO2 emissions (Bárcena Ibarra et al., 2020).
According to OLADE (2020), the share of global CO2 emissions in the
LAC region was about 5.02% while other regions in the world, the figure
was significantly higher than that of the LAC region except for Africa
which accounts for 3.83% of global CO2 emissions (Asia and Australasia
(50.53%), Nort America (16.16%), Middle East (6.33%)).
Fig. 1 shows the changes in the RETI from 2003 to 2019. As can be
seen from the maps, some LAC countries have made a noticeable prog­
ress in terms of renewable energy transition during the period of study
while others have suffered reversal of renewable energy transition. The
former group is represented by Argentina, Brazil, Chile, Colombia, Costa
Rica, Dominican Republic, Honduras, and Paraguay while the latter
group is composed of Bolivia, Jamaica, and Nicaragua.
2.2. Digital economy and its impact on energy sector in the LAC region
When it comes to the digital economy in the LAC region, a growing
Fig. 1. Spatial distribution of the Renewable Energy Transition Index (RETI) in 2003, 2007, 2011, 2015, and 2019.
4
Journal of Cleaner Production 418 (2023) 138146
number of countries in the region have been incorporating digitalization
as a key objective in their policy agenda to boost productivity growth
and energy and material use efficiency in their economy (National
Internet of Things in Brazil, Fourth Industrial Centre in Colombia, digital
manufacturing laboratory in Uruguay) due to the high potential effect
digital economy has on accelerating renewable energy transition,
boosting economic growth, and reducing carbon footprint (CEPAL,
2022b; OECD, 2020). However, despite these efforts made so far, the
progress of digitalization in the LAC region is somewhat slower than
other emerging economies such as Southeast Asian countries and China
and lags far behind that of the industrialized countries (ECLAC, 2022).
For instance, regarding the Internet penetration rate, parameter
frequently used to measure digitalization of economy, was only in 68%
in 2018 (CEPAL, 2022b). Although the number is almost twice
compared to that of 2010, Internet penetration rate in the LAC region is
still far below the OECD average (84%) (OECD, 2020). The lack of
qualified workers capable to manage advanced digital technologies
constitutes another barrier that prevents digitalization in the LAC region
(Jimenez and Gonzalez, 2022). For instance, according to the OECD
statistics in 2018, only one-third of workers in the LAC region use digital
technology related devices (computers, ICT tools, smartphones)
compared to more than half of workers in EU (OECD, 2020). Digital
Y.K. Hwang
economy might offer a great opportunity to accelerate renewable energy
transition and decoupling economic growth from environmental
degradation in the LAC region for the following reasons: First, the
application of advanced digital tools such as AI, machine learning, block
chain in energy sectors can facilitate enormously the integration of
intermittent and variable renewable energy as it ensures a more reliable
electricity supply and enhances energy security and flexibility from
renewable energy sources, thus accelerating low-carbon economy
transition in the region (ECLAC, 2022). However, it is worth noting that
the capabilities of the national energy supply system and the abilities of
workers to assimilate advanced digital technologies are key factors for
the successful integration of renewable energy sectors. Unlike advanced
countries, many Latin American countries face persistent challenges,
such as a lack of skilled workers and inefficient energy systems, to
handle complex technologies like advanced digital technologies and
fully leverage their benefits. Therefore, it is important for governments
to provide strong support for enhancing digital literacy and skills, and
enabling the development of digital capabilities from the outset (ECLAC,
2022). Second, the application of digital technologies in value chain can
significantly improve productivity of workers, energy efficiency and
induce technological innovations as well as structural transformation,
namely shift from energy-intensive economy to knowledge-intensive
and service-based economy (Nagasawa et al., 2017), thus contributing
to economic growth and environmental sustainability through less en­
ergy consumption, technological progress, and cost savings (Ren et al.,
2021). Third, the digital economy can accelerate the development of an
integrated energy system where different sectors of an economy will
become more interconnected and can freely share energy and data
among them, improving the flexibility of national energy system and
generating spillover effects on different sectors (Ren et al., 2021). Also,
digital economy can significantly reduce the frequent mismatch be­
tween supply and demand which is characteristic feature of renewable
energy production (Jimenez and Gonzalez, 2022). Lastly, the LAC is one
of the most urbanized region worldwide with approximately 80% of
population living in the cities which means that the region has high
potential to reap benefits from digitalization as costs saving derived by
improvement in energy efficiency and productivity are expected to be
huge thanks to economies of scale of high urbanization (World Bank,
2022). To take full advantage of the digital economy, issues such as
low-quality network infrastructure, scarcity of qualified workers with
deep knowledge required to manage advanced digital tools in renewable
energy sector, and the high risk of data breaches and hacking in energy
sectors should be addressed in LAC countries (Asif, 2020).
3. Literature review
3.1. Renewable energy, economic growth, and environmental
sustainability
A large number of studies have analyzed the influence of renewable
energy on promoting economic growth in recent years. However, the
empirical results obtained from previous studies so far are not as much
clear as enough to reach a broad consensus about the positive impact of
renewable energy on economic growth. The earlier literatures on the
nexus between renewable energy-economic growth have mainly focused
on how to determine the direction of causality between renewable en­
ergy consumption (REC) and economic growth. Apergis and Payne
(2010) examined the causality between REC and economic growth in 20
OECD countries using FMOLS and cointegration test, and they found
bidirectional causality between them in both the short- and the
long-term. Tugcu et al. (2012) analyzed the impact of REC on economic
growth in G-7 economies using ARDL and they found long-run impact of
REC on economic growth and bidirectional causality between them. The
bidirectional causality between REC and economic growth was
confirmed not only in high-income industrialized countries but also
demonstrated in emerging BRICS countries alike (Sebri and Ben-Salha,
5
Journal of Cleaner Production 418 (2023) 138146
2014). As for confirmation on the positive impact of REC on economic
growth, can also be found in the studies following. Bhattacharya et al.
(2016) used FMOLS, found that renewable energy contributes positively
to economic growth in 57% of 35 top renewable energy consuming
countries in the world. Inglesi-Lotz (2016) used panel cointegration,
pooled estimation, and fixed effect estimation and confirmed the posi­
tive impact of REC (in absolute and relative terms) on economic growth
in 34 OECD countries. Charfeddine and Kahia (2019) found a weak
impact of REC on economic growth and CO2 emissions in 24 MENA
countries on the basis of PVAR estimation. On the other hand, there are
other studies brought about contradictory results. Ocal and Aslan (2013)
estimated the impact of REC on economic growth in Turkey using ARDL
and Toda-Yamamoto Granger causality test and they confirmed a
negative impact of REC on economic growth. The heterogenous impact
of REC on economic growth in terms of country’s income level or sec­
toral composition also has been observed. For example, Ivanovski et al.
(2021) examined the impact of REC on economic growth in both OECD
and non-OECD countries using Local Linear Dummy Variable Estimation
(LLDVE). The estimation results showed that REC had a positive impact
on economic growth in non-OECD countries, but non-significant impact
was found in OECD ones. Sharma et al. (2021b) studied the relationship
between non-renewable and renewable energy consumption and eco­
nomic development in 27 European countries from 1990 to 2016. The
study found a negative relationship between economic growth and
renewable energy consumption, and this relationship was two-way.
Additionally, the study found that the contribution of renewable en­
ergy consumption to economic growth in the 27 EU countries was much
smaller compared to that of non-renewable energy consumption.
Doytch and Narayan (2021) observed that REC is particularly
effective in promoting high-growth sectors in the economy (service
sector in high-income countries and manufacturing sector in
middle-income ones), which further enhanced productivity growth and
consequently led to economic growth of an economy. Wang and Wang
(2020) found a non-linear and positive impact of renewable energy
consumption (as a proxy for renewable energy transition) on economic
growth in OECD countries using panel threshold regression models.
According to the estimation results, when the EU countries were above
the threshold value of renewable energy consumption, its impact on
economic growth was even stronger and more significant. In a similar
vein, Wang et al. (2022a, 2022b, 2022c) investigated the impact of
renewable energy on economic growth in 104 countries using threshold
regression. The estimation results showed positive, non-linear and
heterogenous impacts of renewable energy on economic growth in terms
of income level, resource dependence, and anticorruption regulation.
Regarding the impact of renewable energy on environment, there is a
general consensus among the researchers about the beneficial impact of
renewable energy on improving environmental quality, although the
estimation results can be slightly different depending on the indicators
used as a proxy for environmental quality,2 estimation method, coun­
tries, and period of the study. For example, Dong et al. (2018) have
studied the impact of renewable energy intensity (REI) in 128 countries
and found that it effectively contributed to enhancing environmental
quality by reducing CO2 emissions. Furthermore, the results showed that
the impact of REI was particularly strong in America, Europe, and
Eurasia. Alola et al. (2019) employed Panel Pooled Mean Group ARDL to
examine the impact of REC on environmental quality in 16 EU countries
and they found that REC contributed to improving environmental
quality. Mohsin et al. (2021) used GHG emissions as an indicator of
environmental quality and they observed that REC led to a decrease in
GHG emissions in 25 developing Asian countries. Destek and Sinha
(2020) used FMOLS and DOLS to examine the impact of REC on
ecological footprint in 24 OECD countries. The estimation results
2
the most widely used indicator is CO2 emissions, but GHG emissions and
ecological footprint also have been frequently used.
Y.K. Hwang
showed that ecological footprint decreased as REC increased (thus
positive impact on environmental quality). Sharma et al (2021a)
examined the effect of renewable energy adoption in addition to agri­
culture value added, pesticide use, human capital and economic growth
on greenhosuse gas emissions in the countries of the Bay of Bengal
Instititute for Multisectoral Technical and Economic Cooperation
(BIMSTEC) and found that the interation between renewable energy
adoption and pesticide contributed significantly to mitigating the
negative impact of pesticide use on environmental quality in agriculture
sector. Murshed et al. (2022) investigated the impact of renewable en­
ergy in 25 developing Asian countries using Hausman-Taylor regression
and observed that renewable energy use did not have a direct impact on
carbon productivity. However, energy efficiency gains accompanied by
significant mediation effect between renewable energy use and carbon
productivity could lead to reducing carbon emissions in seven emerging
economies from 2007 to 2018. Trinh et al. (2022) examined the effects
of renewable energy consumption, energy efficiency, and financial
development in mitigating carbon risk across 180 countries from 1980
to 2018. The study revealed that these factors played a substantial role
in significantly reducing carbon risk. Additionally, the authors identified
heterogenous impacts of financial development on the
energy-environment nexus, as well as a U-shaped relationship between
financial development and renewable energy consumption. Dong et al
(2022a) studied the effect of renewable energy development on carbon
emission efficiency in 32 developed countries using Super Efficiency
Slacks-Based Measure (SBM) and Panel threshold regression with fixed
effect. The results of the study implied that renewable energy develop­
ment had a positive impact on carbon emission efficiency, but this
impact came in a non-linear feature, and is significantly different from
one country to another in terms of the threshold value of energy con­
sumption intensity, financial development, renewable energy develop­
ment and carbon emission efficiency of an economy. Likewise, Li et al.
(2022) examined the impact of renewable energy on economic growth
and ecological footprint in 120 countries using panel threshold regres­
sion and found a non-linear impact of renewable energy in terms of
urbanization and income level of country. As can be seen below, the
literature survey on renewable energy-economic growth-environmental
sustainability nexus is summarized in Table 1.
3.2. Digital economy, renewable energy transition, economic growth, and
environmental sustainability
Regarding the literature on the digital economy, recently a growing
body of research has focused on the potential effect it might have on
accelerating renewable energy transition and/or on promoting eco­
nomic growth and enhancing environmental quality in tandem. For
instance, Shahbaz et al. (2022) evaluated the relationship between the
digital economy and energy transition on panel data of 72 countries
during the period 2003–2019. Their estimation showed that the digital
economy effectively contributed to promote renewable energy transi­
tion by reinforcing governance capabilities of governments and the ef­
fect of promotion was larger in high-income countries than in
middle-income countries. And the impact of the digital economy on
energy transition was came out significantly different across the regions
(in Europe and America, the positive effect of the digital economy on
renewable generation was higher while in Asia and the Middle East, its
positive impact was larger on renewable energy consumption). The
investigation of Wang et al. (2022b) in panel data of 72 economies
during the period 2010–2019 showed that the positive impact of the
digital economy was not only limited to just energy transition due to
human capital and financial development but also contributed to
improve the justice of economy such as distributional, procedural, and
restorative justice. Li et al (2022a) analyzed the impact of the digital
economy on green economy efficiency in 281 prefecture-level cities in
China using Slacks Based Measure (SBM) and Spatial Autoregressive
Regression (SAR) and found that the digital economy exerted a positive
6
Journal of Cleaner Production 418 (2023) 138146
effect on green economy efficiency. Moreover, the impact of the digital
economy had an important regional heterogeneity feature (its impact
was greater in the region with high-income level and well-developed
infrastructure and large cities). Liu et al. (2022) investigated the
impact of the digital economy on green total factor productivity in 286
cities in China from 2011 to 2019 and they found a significant positive
impact. Hao et al. (2023) investigated whether digitalization leads to
green economic growth in 30 Chinese provinces and cities during the
period 2013–2019 using System of Environmental and Economic Ac­
counting (SEEA) technique. The estimation results confirmed that digi­
talization had a positive impact on green economic growth through
green technology innovation, advanced industrial structure, and the
rationalization of industrial structure. Moreover, the existence of spatial
effect of the digitalization on green economic growth, regional spatial
heterogeneity, and resource endowment heterogeneity were observed.
Wang et al. (2023) studied the impact of the digital economy on
renewable energy generation (REG) in developed and developing Asian
countries from 2003 to 2019. According to the estimation results ob­
tained using IV-GMM, the impact of the digital economy on REG was
found positive but its impact was especially stronger in developed Asian
economies.
When it comes to the nexus between the digital economy and sus­
tainable development (this is, ensuring economic growth as well as CO2
emissions reductions), there is no consensus among researchers about
whether the digital economy stimulates sustainable development or not.
For instance, Lange et al. (2020) argued that the overall impacts of
digitalization on improving sustainability (reducing energy consump­
tion and thus, energy related CO2 emissions) crucially depended on four
different aspects which were 1) the direct effect of the Information and
Communication Technology (ICT) sector, 2) energy efficiency
improvement thanks to digitalization, 3) energy and labor productivities
increased by digitalization which leads to economic growth and in turn,
generates a growing energy demand and energy consumption, and 4)
structural transformation from energy-intensive to knowledge and
service-oriented economy motivated by wide spread use of digital
technologies and acceleration of digitalization. Authors stated that only
if 2) and 4) dominated over the 1) and 3), digitalization ensured sus­
tainability. In a similar vein, Ren et al. (2021) stated that proliferation of
digitalization in terms of Internet development not only contributed
significantly to increasing energy consumption scale caused by eco­
nomic growth but also helped to reduce energy consumption intensity
and made the energy consumption structure more efficient through
economic growth, human capital and financial development, R&D ac­
tivities, and industrial structure upgrading in 30 Chinese provinces
during the period 2006–2017. On the other hand, Santarius et al. (2020)
investigated the impact of digitalization on decoupling of environmental
degradation from economic growth of 28 European countries, United
States, India, and China during the period 1995–2017 and found that
digitalization on its own did not automatically lead to decoupling,
therefore active political measures along with fundamental changes in
consumption patterns and business models were required to take full
advantages of energy saving effect of digitalization. The study con­
ducted by Ramzan et al. (2022) examined the impact of ICT on
ecological footprint in Pakistan, in addition to financial development,
trade openness, and fossil fuel energy. The results indicated that ICT had
a negative impact on ecological footprint, meaning it increased envi­
ronmental degradation. Moreover, the study found that the interaction
between ICT and both financial development and trade openness further
exacerbated this negative impact on the environment in Pakistan.
However, Xu et al. (2022) in their study of 109 countries from 2000 to
2019, found firm evidence of energy saving effect of digitalization (thus
less energy related emissions) at the international level and this impact
was found to be particularly stronger in low-income developing coun­
tries than high-income developed countries. Similarly, Wang et al.
(2022a) used a digital economy index of 30 Chinese provinces during
the period 2016–2017 to estimate the causal relationship between CO2
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table 1
Literature survey on renewable energy-economic growth-environmental sustainability nexus.
Authors Year Countries Period Estimation methodology Key findings

Apergis & 2010 20 OECD countries 1985–2005 Panel unit root and cointegration tests, FMOLS Presence of short- and long-run bidirectional
Payne causality between REC and economic growth.
Tugcu et al. 2012 G7 economies 1980–2009 ARDL REC had a long-run impact on economic growth.
Bidirectional causality (feedback hypothesis)
between REC and economic growth.
Ocal & Aslan 2013 Turkey 1990–2014 ARDL, Toda-Yamamoto causality test REC had a negative impact on economic growth in
Turkey.
Sebri & Ben- 2014 BRICS countries 1971–2010 VECM, ARDL bounds testing, FMOLS, DOLS Feedback hypothesis between economic growth
Salha and REC
Bhattacharya 2016 38 top renewable energy 1991–2012 FMOLS REC had a significant positive impact on economic
et al. consuming countries in the growth for 57% of total selected countries.
world
Inglesi-Lotz 2016 34 OECD countries 1990–2010 Pedroni cointegration, Pooled estimation, Fixed Both absolute and relative REC contributed positive
Effects estimation and significantly to economic growth in 34 OECD
countries.
Dong et al. 2018 128 countries 1990–2014 CCEMG,AMG REI contributed significantly to reducing CO2
emissions.
The impact of REI was higher in America, Europe,
and Eurasia compared to other regions.
Alola et al. 2019 16 EU countries 1997–2014 Panel Pool Mean Group ARDL REC contributed to improving environmental
quality.
Charfeddine & 2019 24 MENA countries 1980–2015 PVAR REC had a weak impact on economic growth and
Kahia CO2 emissions in 24 MENA countries.
Destek & Sinha 2020 24 OECD countries 1980–2014 FMOLS, DOLS Ecological footprint decreased when REC rises
while NREC led to an increase in ecological
footprint.
A U-shaped relationship between economic growth
and ecological footprint was found and non-
existence of the EKC in OECD countries was
verified.
Wang & Wang 2020 34 OECD countries 2005–2016 Panel threshold regression REC had a positive effect on economic growth.
Existence of nonlinear effect of REC on economic
growth.
Doytch & 2021 107 countries 1984–2019 Dynamic panel GMM, System GMM, REC was effective in promoting growth in high-
Narayan Endogenous growth framework growth sectors (service sectors in high-income
countries while in middle-income ones,
manufacturing sectors).
Complementarity between REC and NREC in high-
income countries while in middle-income
countries, they were substitutes each other.
Ivanovski et al. 2021 OECD and non-OECD 1990–2015 Local linear dummy variable estimation REC had a positive effect on economic growth in
countries non-OECD countries but not in OECD ones.
Mohsin et al. 2021 25 developing Asian 2000–2016 Random effects, Hausman-Taylor regression, A 1% increase in REC contributed to reducing GHG
countries ECM emissions by about 0,193%.
Feedback hypothesis between REC and economic
growth.
Economic growth and REC were positively
correlated in both the short- and the long-run.
Sharma et al. 2021 The countries of BIMSTEC 1985–2019 Second generation unit root tests, Panel An U-shaped relationship between agriculture
a (Bangladesh, India, cointegration, Panel quantile regression value added and GHG emissions.
Myanmar, Nepal, Sri Lanka, The interaction between renewable energy
and Thailand) adoption and pesticide use contributed to
mitigating the negative effect of pesticide use on
environmental quality in agriculture sector.
Sharma et al. 2021 27 European countries 1990–2016 Arellano-Bond dynamic panel data estimation, Two-way positive relationship between economic
b System dynamic panel data estimation, AMG, growth and non-renewable energy consumption.
Quantile regression Two-way negative relationship between economic
growth and renewable energy consumption.
The contribution of non-renewable energy
consumption was much greater than that of
renewable energy consumption.
Positive impact of non-renewable energy
consumption on GDPin all quantile groups except
for Q90 in quantile regression.
Dong et al. 2022 32 developed countries 2000–2018 Super-efficiency slacks-based measure, Panel Renewable energy development led to an
a threshold model with interactive fixed effects improvement in carbon emission efficiency.
However, the impact of renewable energy
development on carbon emission efficiency was
non-linear and differs significantly in terms of
energy consumption intensity, financial
development, renewable energy development, and
carbon emission efficiency level of country.
Li et al. 2022 120 countries 1995–2014 Panel threshold regression model, Fixed Effect Non-linear impact of renewable energy on
b model economic growth and ecological footprint
(continued on next page)

7
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table 1 (continued )
Authors Year Countries Period Estimation methodology Key findings

depending on countries’ urbanization and income

level.
The negative impact of renewable energy on the
ecological footprint reduced and then increased
when urbanization exceeded the threshold value.
Renewable energy had a positive effect on
economic growth and its effect became larger as
urbanization level increased.
Murshed et al. 2022 7 emerging countries 2007–2018 Mediation regression model Energy efficiency exerted important mediation
effect between carbon productivity and renewable
energy use to reduce carbon emission levels in 7
emerging countries. Renewable energy did not have
direct impact on carbon productivity.
Trinh et al. 2022 180 countries 1980–2018 OLS, Robust standard errors methods, FE panel REC and improved energy efficiency effectively
methods, GMM (two-step GMM, system GMM), con- tributed to mitigating climate risk.
Quantile regression methods (Canay’s method, Inverted U-shape relationship between financial
Powell’s method, Machado and Silva’s method) development and REC.
Heterogenous impacts of financial development on
the energy-environment nexus.
Wang et al. 2022 104 countries 2002–2018 FOLS, Threshold regression Positive relationship between renewable energy
and economic growth.
The impact of renewable energy on economic
growth in 3 different groups (high-income, middle-
income, and low-income groups) varied greatly in
terms of resource dependence and anticorruption
regulation.

Note: AMG: Augmented Mean Group; ARDL: Autoregressive Distributed Lag; BIMSTEC: Bay of Bengal Initiative for Multisectoral Technical and Economic Cooperation;
BRICS: Brazil, Russia, India, China, and South Africa; CCEMG: Common Correlated Effects Mean Group DOLS: Dynamic Ordinary Least Squares; ECM: Error Correction
Model; EKC: Environmental Kuznets Curve; EU: European Union; FMOLS: Fully Modified Ordinary Least Squares; GHG: greenhouse gases; GMM: Generalized Method
of Moments; MENA: Middle east and North Africa; NEC: nuclear energy consumption; NREC: non-renewable energy consumption; OECD: Organization for Economic
Cooperation and Development; REC: renewable energy consumption; REI: renewable energy intensity; VECM: Vector Error Correction Model.

emissions and the digital economy using the System GMM technique.
They found that the digital economy negatively affected CO2 emissions
via expansion of tertiary sector which reduced the share of coal con­
sumption and promoted green technology innovation as well. Dong et al
(2022b) examined the effect of information infrastructure on urban GHG
emissions in 281 prefecture-level cities in China using Difference in
Difference (DID) analysis. They found that technological innovation,
factor allocation enhancement, and tertiary agglomeration were three
major factors through which information infrastructure contributed to
reducing GHG emissions in China. Zhang et al. (2022) studied the
relationship between the digital economy and carbon emission perfor­
mance (CEP) in 277 cities in China and found that digital economy
improved CEP, but its impact was heterogenous and non-linear in Chi­
nese cities. Ma et al. (2022) analyzed the impact of digitalization on the
provincial emission levels in 30 Chinese provinces using Augmented
Mean Group (AMG) and Common Correlated Effect Mean Group
(CCEMG) estimators. The estimation results indicated that digitalization
effectively contributed to reducing emission levels in Chinese provinces.
The existence of significant moderating effects of R&D investments and
technological innovation between digitalization and CO2 emissions
were also confirmed. Lastly, Zhang (2023) investigated the threshold
effect of digital transformation on carbon emissions using the panel data
of 29 major exporting countries during the period 2000–2019 to
examine the impact of energy consumption on CO2 emissions. According
to their study, when digital transformation went beyond the threshold
value, the positive impact of energy consumption on environmental
degradation (increases in carbon emissions) began to decrease and the
promoting effect of renewable energy on energy saving and carbon
emission reduction became larger. As can be seen below, the literature
survey on digital economy-economic growth-environmental sustain­
ability nexus is summarized in Table 2.
From the literature review, we came to recognize that there are
significant gaps in the extant research. Firstly, regarding the renewable
energy, most previous studies have used unidimensional variables of
demand or supply on either side, such as renewable energy consumption
or renewable energy generation, to assess the impact of renewable en­
ergy transition on economic growth and environmental sustainability.
Only a few studies have developed multidimensional index of energy
transition that accounts for the various facets of complex energy trans­
formation process such as political, institutional, social, economic as­
pects. Secondly, the LAC region lacks research on the synergistic effect
between the digital economy and renewable energy transition. Most
previous studies in this research have focused on emerging Asian
countries, particularly China (either at the provincial or prefecture
level), or industrialized countries (such as EU member countries, the
USA, and Canada), to examine the impact of the digital economy on
accelerating decarbonization and the decoupling of economic growth
from environmental degradation Finally, the extant studies on the
impact of renewable energy in the LAC region have failed in considering
the heterogeneous and non-linear nature of the renewable energy
transition and the digital economy on economic growth and environ­
mental sustainability. The previous studies have mostly used standard
linear regression models to analyze this impact, which are unable to
capture non-linearity and regional heterogeneity and may bring about
biased estimates. Having identified the gaps in extant research through
literature review and extensive survey, three different hypotheses are
formulated as follows.
Hypothesis 1. Renewable energy transition, when combined with the
digital economy, it might bring about synergistic effects and thus
contribute greatly to accelerating economic growth and improving
environmental quality alike in terms of CO2 emissions reduction in the
LAC region.
Hypothesis 2. The impact of transition to renewable energy on eco­
nomic growth and environmental quality implies a significant threshold
effect in the LAC region. In other words, depending on whether the

8
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table 2
Literature survey on digital economy-economic growth-environmental sustainability nexus.
Authors Year Countries Period Estimation methodology Key findings

Lange 2020 EU countries, Norway, Switzerland, 1995–2016 Theoretical formulation based on Digitalization did not contribute to decoupling of
et al. Australia, Brazil, Canada, China, Brock and Taylor (2005) economic growth form energy consumption.
India, Japan, South Korea, Russia, ICT sector led to an increase in energy consumption.
Taiwan, USA
Santarius 2020 28 EU countries, US, India, China 1995–2017 Theoretical approach Digitalization on its own did not automatically lead to a
et al. decoupling of economic growth from environmental
degradation.
Ren et al. 2021 30 provinces in China 2006–2007 OLS, System-GMM Positive correlation between China’s energy consumption
and Internet development.
Internet development could promote the energy
consumption scale via economic growth, but it could also
contribute to reducing energy consumption intensity
through channels of economic growth, R&D investments,
financial development, industrial structural upgrading,
and human capital.
Important regional differences regarding the effect of
Internet development on energy consumption scale,
structure, and intensity.
Dong et al. 2022 281 prefecture-level cities in China 2003–2018 DID analysis Information infrastructure had a significant effect on
b reducing urban greenhouse gases emission levels.
Technological innovation, industrial structural
upgrading, factor allocation enhancement, and tertiary
agglomeration were the mechanisms through which
information infrastructure improved environmental
performance.
Li et al. 2022 277 cities in China 2011–2018 SBM,SAR Digital economy had a significant positive impact on
a green economy efficiency.
Impact of the digital economy differed significantly
across the regions. Stronger impact in the Eastern region
and large cities while in the Central, Western regions and
small cities, the impact of the digital economy was
smaller.
Liu et al. 2022 286 cities in China 2011–2019 DDF, GML, Tobit, Quantile Digital economy improved significantly green total factor
regression, Impulse response productivity.
function, Intermediary effect Important regional heterogeneity observed.
model Digital economy promoted green total factor productivity
through industrial structure upgrading.
Ma et al. 2022 30 Chinese provinces 2006–2017 AMG,CCEMG Digitalization contributed to curbing the provincial
emission levels.
Significant moderating effect of R&D investments and
technological innovation between digitalization and CO2
emissions.
Economic growth, financial development, and energy use
contributed significantly to increasing CO2 emissions.
Ramzan 2022 Pakistan 1960 Q1 to Non-parametric causality in The interaction between ICT and FID increased
et al. 2019 Q4 quantile techniques, Diks and significantly the ecological footprint.
Panchenko nonlinear Granger The interaction between ICT and trade openness
causality test increased significantly the ecological footprint.
ICT, financial development and fossil fuel energy had a
significant impact on ecological footprint.
Shahbaz 2022 72 countries 2003–2019 System-GMM, Panel quantile Digital economy had a positive impact on energy
et al. regression transition (in terms of both REC and REG).
Governments’ governance capabilities played a key role
in promoting energy transition through digital economy.
Strong impact of the digital economy on energy transition
at higher quantiles.
Existence of regional heterogeneities regarding the
impact of the digital economy on energy transition.
Wang et al. 2022 30 Provinces in China 2006–2017 System-GMM Digital economy had a negative effect on CO2 emissions.
a The infrastructure, innovation and application, economic
growth, and jobs of the digital economy contributed to
reducing CO2 emissions.
Wang et al. 2022 72 countries 2010–2019 System-GMM, Mediating effect Digital economy promoted a just transition.
b model Human capital and financial development were the two
mechanisms through which the digital economy
indirectly enhanced just transition.
Xu et al. 2022 109 Countries in Asia Pacific, Europe, 2000–2019 System-GMM, Mediating effect Digitalization contributed to reducing energy
Africa model consumption, energy intensity, and optimizing energy
structure.
Technological innovation, human capital, and industrial
structure were the three main mechanisms through which
the digitalization indirectly affects energy.
In low-income and developing countries, digitalization
had a greater impact on the energy.
(continued on next page)

9
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table 2 (continued )
Authors Year Countries Period Estimation methodology Key findings

Zhang 2022 277 cities in China 2011–2019 EBM, SDM, OLS, Mediation effect Digital economy enhanced carbon emission performance.
et al. model, Threshold regression model The main mechanisms through which the digital
economy affected carbon emission performance were
energy intensity, energy consumption scale, and urban
afforestation.
Non-linear and different impact of the digital economy on
carbon emission performance depending on the levels of
energy intensity, energy consumption structure,
government intervention, and urban afforestation.
Existence of spatial effect of the digital economy on
carbon emission performance.
Hao et al. 2023 30 Chinese provinces and cities 2013–2019 SEEA Digitalization significantly led to green economic growth.
Green technology innovation, advanced industrial
structure and the rationalization of industrial structure
have important mediation effects between digitalization
and green economic growth.
The existence of positive local and neighboring effect of
digitalization on green economic growth.
The existence of regional spatial heterogeneity and
resource endowment heterogeneity.
Wang et al. 2023 Developed and developing Asian 2003–2019 IV-GMM Digital economy had a positive impact on REG.
countries The impact was particularly strong in developed
economies in Asia.
Existence of regional heterogeneity as the impact of the
digital economy on REG was positive and significant only
in East and South Asian countries.
Zhang 2023 29 major exporting countries in the 2000–2019 Multivariate threshold regression High digitalization contributed to reducing the
et al. world promotion effect of per capita energy consumption on
carbon emissions.
The promotion effect of renewable energy on energy
conservation and carbon emission reductions increased
when digital infrastructure development, digital trade
competitiveness, and digital technology exceeded each
threshold value.

Note: DDF: Direction Distance Function; DFI: Digital Financial Inclusion; DID: Difference In Difference; EBM: Epsilon Based Measure; EEP: Energy-Environment
Performance; FID: Financial Development; ICT: Information and Communication Technologies; GML: Global Marmquist-Luenberger; IV-GMM: Instrumental Vari­
able- Generalized Method of Moments; REG: Renewable Energy Generation; R&D: Research and Development; SAR: Spatial Autoregressive; SBM: Slacks Based
Measure; SDM: Spatial Durbin Model; SEEA: System of Environmental and Economic Accounting.

countries are below or above the threshold value,3 its impact will be very
different.
Hypothesis 3. The impact of renewable energy transition and the
digital economy on economic growth and environmental quality shows
a significant heterogeneity across the quantile distribution. In other
words, depending on whether countries belong to one quantile or
another, the impact of renewable energy transition on economic growth
and environmental quality will be significantly different in the LAC
region.
4. Methodology and data
In this section, we provide both empirical and theoretical framework
on which our analysis is based.
4.1. Principal factor analysis (PFA) to elaborate renewable energy
transition index (RETI) and digital economy index (DEI)
As principal objectives of this investigation are focused on estimating
the impact of renewable energy transition on economic growth as well
as environmental degradation in the LAC countries and analyzing the
synergistic effect of the digital economy on renewable energy transition
from the perspective of economic growth and environmental
degradation, the Principal Factor Analysis (PFA) is used to elaborate
both the Renewable Energy Transition Index (RETI) and Digital Econ­
omy Index (DEI). The main reason why we use the composite index
instead of the single-dimension indicator is that both renewable energy
transition4 and digital economy imply multidimensional and complex
aspects such as economic,social, political and technical components, so,
using one single-dimension indicator cannot reflect well the complex
reality of the energy transition and the digital economy.
4.1.1. Renewable energy transition index (RETI)
Eleven indicators are selected to construct the RETI based on the data
availability and previous studies by Kuc- Czarnecka et al., (2021), Singh
et al. (2019), and Višković et al. (2022). Prior to application of the PFA,
the stationarity and the normalization or the standardization of the in­
dicators should be done to avoid spurious results and to ensure that all of
the indicators are at the same scale. The stationarity of each indicator
used in the RETI is examined with the unit root test and the test results
indicate that all of them are stationary, therefore only the standardiza­
tion of the data is required to be done. Based on the results of the PFA,
four factors identified are grouped into four dimensions to construct the
RETI. The weights of each dimension are assigned based on the eigen­
values obtained from the PFA. (see Appendix). Table 3 illustrates the
RETI and its four principal dimensions.

4
In previous studies, both renewable energy consumption and renewable
3
In this study, both per capita income and per capita CO2 emissions are used energy supply are more frequently used variables to capture the progress of
as threshold variables. energy transition.

10
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table 3
Renewable energy transition index (RETI).
Renewable Energy Transition Index

Dimension Indicator (factor) Unit Data source Attribute

Institutions, regulations, and governance Control of Corruption Index (factor 1) World Bank +
(41%) Government Effectiveness Index World Bank +
(factor 1)
Regulatory Quality Index (factor 1) World Bank +
Investment freedom index (factor 1) The Global Economy.com +
Energy system structure (17%) RE share of electricity generation % Of total electricity IRENA +
(factor 4) generation
Manufacturing, value added (factor % Of GDP World Bank +
4)
Energy use (Energy efficiency) (factor $2017 PPP GDP/MJ British Petroleum (BP), Energy Information +
4) Administration (EIA)
Universal energy access, facility to start-up Access to electricity, rural (factor 3) % Of rural population WB +
new business (20%) Financial freedom index (factor 3) The Global Economy.com +
Human capital, large-scale projects funding Scientific and technical journal % Of GDP World Bank +
opportunities (22%) articles (factor 2)
Bank credit to the public sector % Of GDP World Bank +
(factor 2)

4.1.2. Digital economy index (DEI)
Developing the DEI is based on a set of 13 indicators in consideration
of data availability and previous studies by Muhammad Shahbaz et al.
(2022), Wang et al. (2022a), and Wang et al. (2022b). Most indicators
used in constructing the DEI are non-stationary ones unlike those used in
constructing the RETI. Therefore, they are first differenced and
normalized before conducting the PFA. A total of six factors is identified
and then reduced to three dimensions for their use in constructing the
DEI (see in Table A.0 in Appendix). Table 4 shows the DEI and its three
dimensions.
4.2. Control variables
To mitigate potential omitted variable bias, the following variables
have been chosen as control variables.
• Total natural resource rents: Many Latin American economies’ fiscal
capacity and gross domestic product heavily rely on revenues ob­
tained from the extraction of non-renewable natural resources: oil,
natural gas, and critical minerals required for low-carbon energy
transition, such as copper, lithium, and cobalt (CEPAL,N 2022 a).
However, this dependence on natural resource exploitation can also
lead to significant environmental costs as the extractive industries in
the region are often energy-intensive, and emit high levels of CO2,
contributing to increased carbon emissions in the LAC region
(Alvarado et al., 2021).
• Inflation level:The economies in the LAC region have been experi­
encing persistent inflationary pressures. From an economic growth
standpoint, high levels of inflation can result in significant negative
effects on household purchasing power, production costs, and create
uncertainty of future levels of price, which can undermine in­
vestments and savings (CEPAL,N 2022 a). From an environmental
perspective, high levels of inflation can influence consumption
behavior, and lead individuals and households to opt for cheaper and
less sustainable energy sources. For example, traditional biomass
energy, which may be more affordable during periods of high
inflation, can aggravate environmental degradation and pollution
(Urban, 2014). Additionally, high levels of inflation can impact

Table 4
Digital economy index (DEI)a.
Digital Economy Index

Dimension Indicator (factor)b Unit Data source Attribute

Digital Economy Infrastructure (58%) E-Participation Index (factor 1) United Nations (UN) +
Online Service Index (factor 1) United Nations (UN) +
Mobile cellular subscriptions (factor 2) per 100 people International Telecommunication +
Union (ITU)
Fixed broadband subscriptions (factor 2) per 100 people International Telecommunication +
Union (ITU)
Services, value added per worker (factor constant 2015 US$ World Bank (WB) +
2)
Fixed telephone subscriptions (factor 3) per 100 people International Telecommunication +
Union (ITU)
Telecommunication Infrastructure Index United Nations (UN) +
(factor 3)
Quality of workers & possession of communication Human Capital Index (factor 5) % Of total population International Telecommunication +
appliances (26%) Union (ITU)
Internet Users (factor 6) United Nations (UN) +
Wage and salaried workers (factor 6) % Of total World Bank (WB) +
employment
Services, value added (factor 6) % Of GDP World Bank (WB) +
Technology-intensive goods trade (16%) ICT goods exports (factor 4) % Of total goods World Bank (WB) +
exports
ICT goods imports (factor 4) % Of total goods World Bank (WB) +
imports
a
Before performing Principal Factor Analysis, each components of the DEI are first differenced and normalized.
b
Factor refers to the factor assigned during the PCF analysis.

11
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

government fiscal capacity and restrict the availability of public Table 5

funds for environmental protection efforts. Variable description.
• Foreign Direct Investment (FDI): FDI might have an ambiguous effect Variable Explanation Unit Data Source
on both economic growth and environmental sustainability in the
GDPpc Per capita Gross Domestic constant World Bank
LAC region. On the one hand, FDI can facilitate the transfer of Production. Proxy for economic 2015 (US$) (WB)
advanced technologies and managerial expertise, which can enhance growth
productivity and stimulate economic growth in the region (Ben Jebli CO2pc Per capita CO2 emissions. Proxy for metric tons World Bank
et al., 2019). On the other hand, FDI can also attract investments in environmental degradation or per capita (WB)
environmental quality
natural resource exploitation and energy-intensive industries RETI Renewable Energy Transition Index – Various
(Doytch and Narayan, 2016). This can result in the lock-in of DEI Digital Economy Index – Various
carbon-intensive technologies and practices, and lead to negative RETIxDEI Interaction term between Renewable – Various
environmental impacts. The focus on resource extraction and Energy Transition Index and Digital
Economy Index
energy-intensive sectors may cause increased pollution, deforesta­
TNRR Total Natural Resource Rents % Of GDP World Bank
tion, and carbon emissions, thereby posing challenges to environ­ (WB)
mental sustainability in the LAC region. Inflation Inflation, GDP deflator annual % World Bank
• Globalisation: Globalisation is another variable to consider when (WB)
examining economic growth and environmental degradation. The FDI Foreign Direct Investment, net % Of GDP World Bank
inflows (WB)
process of globalisation in the LAC region has been accelerating since GI Globalization Index – The Global
the Washington Consensus in 1992 and brought about significant Economy.
changes in the economic structure of many LAC countries, including com
privatization and market liberalization (Santiago et al., 2020). From AFFVadd Agriculture, Forestry, and Fishing % of GDP World Bank
Value added (WB)
an economic growth perspective, globalisation has been facilitating
greater trade and investment flows, as well as the integration of
many Latin American countries into the global markets. However, it
has exposed many economies in the region to global competitiveness Table 6
and increased vulnerability to external shocks as well. From an Descriptive statistics.
environmental perspective, increased economic activities and trade Variable Obs Mean Std. dev. Min Max
might lead to higher energy consumption and CO2 emissions, LnGDPpc 306 8.661 0.6131 7.2944 9.6827
thereby worsening environmental quality. However, it is important LnCO2pc 306 0.5806 0.5335 − 0.4778 1.573
to note that globalisation can also facilitate the transfer of cleaner LnRETI 306 − 1.0853 0.3574 − 2.0480 − 0.4947
technologies and knowledge transfers, which can contribute to LnDEI 306 − 0.7314 0.1330 − 1.2987 − 0.1361
LnRETIxLnDEIa 306 0.0099 0.0430 − 0.1499 0.2558
environmental improvement. LnTNRR 306 0.7943 1.1333 − 2.3308 2.937
• Agriculture, Forestry, and Fishing Value added: Unlike other regions LnInflation 306 2.3224 0.5209 − 2.5333 4.0186
in the world, a significant share of greenhouse gas (GHG) emissions LnFDI 306 1.7651 0.6381 − 6.7944 2.9301
in the LAC region stems from agricultural production and land use LnGI 306 4.1669 0.0853 3.8979 4.362
LnAFFVadd 306 1.9393 0.4667 0.7825 2.9384
change, such as the expansion of agricultural lands and deforestation
(OLADE, 2020). Specifically, in Latin America, 42.2% of the total a
To avoid issues of multicollinearity in our estimation, the interaction term is
GHG emissions come from agriculture, livestock, land use change, previously centered. This involves subtracting the mean values of both LnRETI
and forestry, whereas these sectors account for only 17.4% of the and LnDEI from their respective variables. Then, LnRETI and LnDEI are multi­
total GHG emissions globally (Bárcena Ibarra et al., 2020). plied to obtain the interaction term, LnRETIxLnDEI.

4.3. Variables and data descriptions
In this study, a balanced panel of 18 Latin American and Caribbean
countries during the period 2003–2019 is used, namely Argentina,
Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic,
Ecuador, El Salvador, Guatemala, Honduras, Jamaica, Mexico,
Nicaragua, Panama, Paraguay, Peru, and Uruguay, to investigate the
impact of transition to renewable energy on economic growth and
environmental degradation and the synergistic effect of the digital
economy on renewable energy transition from the perspective of eco­
nomic growth and environmental degradation. For this purpose, a set of
10 variables are used (see Table 5). The key independent variable are the
Renewable Energy Transition Index (RETI), the Digital Economy Index
(DEI), and the interaction term between the RETI and DEI (RETIxDEI).
Before proceeding to the analysis, all variables are transformed into
logarithmic form to deal with Skewness and make them conform to
normality (see Table 6). Furthermore, the log-transformation of the
variables will facilitate the interpretation of the coefficients in regres­
sion model in terms of elasticity. Before conducting the analysis, it is
worth noting that the LAC region has been facing three serious issues of
data use: availability, quality and quality. The national data of most
countries in the LAC region required for this research employ the System
of National Accounts (SNA, 1993) standard for measuring GDP. How­
ever, this standard does not properly take into account the informal
economy, which is prevalent in the LAC region. To alleviate the issues of
data facing this study: availability, quality and quantity, the World Bank
database was used for this research.
4.4. Preliminary tests
Before proceeding the linear- and non-linear regression, a set of
preliminary tests is required to examine issues of correlation and mul­
ticollinearity, cross sectional dependence, the presence of unit root,
cointegration relationships among the variables, and the existence of
fixed or random effects in regression model (Fuinhas and Marques,
2019). Furthermore, when applying the Methods of Moments Quantile
Regression (MMQR), Swamey, Shapiro-Wilk and Skewness-Kurtosis test
are conducted to examine whether the slope homogeneity and the re­
siduals of the dependent variables in regression models, namely
LnGDPpc and LnCO2pc, follow a normal distribution or not (Fuinhas
and Marques, 2019; Machado and Silva, 2019). In case of the
Fixed-Effect Panel Threshold Regression (FEPTR), the threshold test is
previously done to analyze the existence of meaningful threshold points
and thus the applicability of the FEPTR.
4.5. Standard regression models (Pooled Ordinary Least Squares (POLS),
random effects (RE) estimator, Driscoll-Kraay Fixed Effects (D-K FE)
estimator, Fixed Effects Two Stage Least Squares (FE-2SLS))
The standard linear regressions, namely POLS, RE, the D-K FE, and

12
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table 7
Standard linear regressions.
Dependent variable: LnGDPpc

Estimation POLS RE Driscoll-Kraay FE FE-2SLS

technique
Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2

LnTNRR − 0.013 (0.022) − 0.010 (0.022) − 0.047*** − 0.047*** − 0.046*** − 0.046*** − 0.060*** − 0.062***
(0.014) (0.014) (0.010) (0.011) (0.015) (0.015)
LnInflation 0.206*** 0.204*** − 0.055*** − 0.055*** − 0.059** − 0.058** − 0.055*** − 0.053***
(0.046) (0.046) (0.015) (0.015) (0.026) (0.025) (0.015) (0.015)
LnFDI − 0.102 (0.041) − 0.104** − 0.002 (0.012) − 0.002 (0.012) − 0.003 (0.017) − 0.003 (0.017) − 0.009 (0.012) − 0.008 (0.012)
(0.041)
LnGI 4.012*** 4.070*** 2.144*** 2.141*** 2.099*** 2.088*** 2.163*** 2.145***
(0.339) (0.341) (0.149) (0.152) (0.129) (0.133) (0.188) (0.187)
LnRETI 0.336*** 0.334*** 0.022 (0.047) 0.022 (0.047) 0.006 (0.034) 0.004 (0.033) ¡0.037 ¡0.046
(0.082) (0.081) (0.062) (0.062)
LnDEI 0.800*** 0.796*** 0.115** 0.116** 0.110* (0.056) 0.110* (0.056) 0.412** 0.399**
(0.187) (0.187) (0.049) (0.050) (0.178) (0.174)
LnRETIxLnDEI 0.742 (0.568) ¡0.053 ¡0.069 ¡0.380**
(0.151) (0.213) (0.179)
Constant − 7.395*** − 7.644*** 0.003 (0.649) 0.018 (0.659) 0.177 (0.610) 0.220 (0.632) 0.101 (0.847) 0.159 (0.840)
(1.473) (1.483)
Country FE No No No No Yes Yes Yes Yes
Year FE No No No No No No No No
R2 0.540 0.543 0.405 0.405 0.396 0.395 0.410 0.405
Number of obs. 306 306 306 306 306 306 288 288

Dependent variable: LnCO2pc

Estimation POLS RE Driscoll-Kraay FE FE-2SLS

technique
Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4

LnAFFVadd − 0.974*** − 0.976*** − 0.347*** − 0.347*** − 0.286*** − 0.290*** − 0.246*** − 0.252***

(0.051) (0.051) (0.052) (0.052) (0.069) (0.068) (0.057) (0.057)
LnInflation 0.174*** 0.175*** − 0.039** − 0.038** − 0.046** − 0.045** − 0.048*** − 0.047***
(0.033) (0.033) (0.017) (0.017) (0.018) (0.018) (0.017) (0.017)
LnFDI 0.003 (0.028) 0.006 (0.028) 0.034** (0.014) 0.034** (0.014) 0.034*** 0.034** (0.034) 0.032** 0.033**
(0.006) (0.014) (0.014)
LnGI 1.599*** 1.542*** 0.578*** 0.540*** 0.580*** 0.543*** 0.607** 0.576**
(0.258) (0.259) (0.188) (0.191) (0.105) (0.121) (0.235) (0.235)
LnRETI ¡0.638*** ¡0.636*** ¡0.145*** ¡0.148*** ¡0.140** ¡0.144*** ¡0.130* ¡0.139*
(0.064) (0.064) (0.054) (0.054) (0.048) (0.048) (0.073) (0.073)
LnDEI 0.228* (0.132) 0.230* (0.132) 0.079 (0.058) 0.079 (0.057) 0.073 (0.052) 0.073 (0.054) 0.224 (0.197) 0.209 (0.192)
LnRETIxLnDEI ¡0.651 (0.398) ¡0.186 (0.175) ¡0.178 (0.211) ¡0.359*
(0.205)
Constant − 5.130*** − 4.887*** − 1.224 (0.853) − 1.070 (0.863) − 1.333** − 1.177* − 1.390 (1.098) − 1.271 (1.101)
(1.136) (1.143) (0.512) (0.558)
Country FE No No No No Yes Yes Yes Yes
Year FE No No No No No No No No
R2 0.698 0.700 0.593 0.598 0.574 0.582 0.559 0.578
Number of obs. 306 306 306 306 306 306 288 288

Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively; Model 1–4 correspond to equation (1)-(4) respectively in section 4.4. The values in
parentheses indicate the standard error.

FE-2SLS estimators are performed before the non-linear regression

models (the FEPTR and the MMQR) to detect the effect of transition to
renewable energy as well as the interaction term between the RETI and
DEI on economic growth and environmental degradation. The results
obtained from the standard linear regression models will be compared
with those of the MMQR and FEPTR models to see whether the estimated
coefficients effectively show the non-linear and heterogenous behaviors
in our panel data.
The benchmark regression models are constructed as follows:
LnGDPpcit = b10 + b11 LnTNRRit + b12 LnInflationit + b13 LnFDIit + b14 LnGIit
+ c11 LnRETIit + c12 LnDEIit +a1i + e1it
(1)
LnGDPpcit =b20 + b21 LnTNRRit + b22 LnInflationit +b23 LnFDIit +b24 LnGIit
+c21 LnRETIit +c22 LnDEIit + c23 (LnRETI ∗ LnDEI)it + a2i + e2it
(2)
When the LnCO2pc is considered as an outcome variable:
LnCO2pcit = b30 + b31 LnAFFVaddit + b32 LnInflationit + b33 LnFDIit
+ b34 LnGIit + c31 LnRETIit + c32 LnDEIit + a3i + e3it (3)
LnCO2pcit = b40 + b41 LnAFFVaddit + b42 LnInflationit + b43 LnFDIit
+ b44 LnGIit + c41 LnRETIit + c42 LnDEIit (4)
+ c43 (LnRETI ∗ LnDEI)it + a4i + e4it
Where the subscript i denotes the country, t represents year, bk0 , bk1 , bk2 ,
bk3 , and bk4 represent a constant and the coefficients of control variables
respectively, ck1 , ck2 , and ck3 represent coefficients of the independent
variables LnRETI, LnDEI, and LnRETIxLnDEI respectively, aki denotes
country fixed effects, and ekit represents the error term. The variables
LnGDPpc, LnCO2pc, LnTNRR, LnInflation, LnFDI, LnGI, LnAFFVadd,
LnRETI,LnDEI, and LnRETIxLnDEI denote per capita gross domestic
product, per capita carbon dioxide emissions, total natural resource
rents, inflation level, foreign direct investment, globalization index,
agriculture, forest, and fishery value added, renewable energy transition
index, digital economy index, and the interaction term between the
renewable energy transition index and the digital economy index

13
Y.K. Hwang
respectively (synergistic effect between renewable energy transition and
the digital economy). The interaction term is incorporated in regression
models to see whether the digital economy has a meaningful synergistic
effect with renewable energy transition on promoting economic growth
and improving environmental sustainability by reducing CO2 emissions.
It is worth noting that to obtain LnRETIxLnDEI, both LnRETI and LnDEI
are previously centralized (demeaned), and then multiplied them to
finally obtain the interaction term. In this way, the potential multi­
collinearity issue can be addressed. Finally, all variables are transformed
into logarithmic forms to reduce the variability and to facilitate the
interpretation of the coefficients estimated in terms of elasticities.
Both coefficients c23 and c43 are required to evaluate the synergistic
effect. If the parameter c23 in equation (2) is positive and statistically
significant, the synergistic effect on economic growth exists. Similarly, if
the parameter c43 in equation (4) is negative and statistically significant,
this means that the digital economy combined with renewable energy
transition effectively contributes to enhancing environmental quality.
4.6. FixedEffect Panel Threshold Regression (FEPTR)
The FEPTR is used to analyze whether the impacts of energy tran­
sition on economic growth and environmental degradation are different
and show non-linearity and asymmetry depending on whether country’s
per capita income or per capita CO2 emissions are above or below the
threshold values.
The FEPTR model with a single threshold can be formulated as fol­
lows:
∑
Yit = μi + δ1 Rit ∗ I(Qit ≤ γ 1 ) + δ2 Rit ∗ I(Qit > γ 1 ) + θm Zit + Vt + εit (5)
In case of multiple thresholds:
Yit = μi + δ1 Rit ∗ I(Qit ≤ γ 1 ) + δ2 Rit ∗ I(Qit > γ 1 ) + δn Rit
∑ within countries when estimating the coefficients.
m
∗ I(Qit ≥ γk )…. + θm Zit + Vt + εit
n=1
Where μi and Vt denote a country and time fixed effect, Rit , Zit , and , Qit
denote a regime-dependent, independent, and threshold variable
respectively, I (.) represents an indicative function, θi and δi denote the
coefficients of the independent and the regime-dependent variables in
different intervals respectively (below and above the threshold value), γk
represents a critical value of threshold variable Qit , and εki is the inde­
pendent and identically distributed error term.
Adapting the FEPTR to our regression models, the FEPTR can be
formulated as follows:
LnGDPpcit = α1i + θ11 LnTNRRit + θ12 LnInflationit + θ13 LnFDIit + θ14 LnGIit
+ θ15 LnDEIit + θ16 LnRETIxLnDEIit + δ11 LnRETIit ∗ I(Qit ≤ γ 1 )
+ δ12 LnRETIit ∗ I(Qit > γ1 ) + V1t + ε1it
LnGDPpcit = α2i +θ21 LnTNRRit +θ22 LnInflationit +θ23 LnFDIit +θ24 LnGIit
+θ25 LnDEIit +θ26 LnRETIxLnDEIit +δ21 Ln RETIit ∗I(Qit ≤γ 1 )
+δ22 Ln RETIit ∗I(γ1 <Qit ≤γ2 )+δ23 LnRETIit ∗I(LnGDPpcit >γ2 )
+V2t + ε2it
When the LnCO2pc is considered as an outcome variable:
LnCO2pcit = α3i + θ31 LnAFFVaddit + θ32 LnInflationit + θ33 LnFDIit
+ θ34 LnGIit +θ35 LnDEIit + θ36 LnRETIxLnDEIit + δ31 LnRETIit
( ) ( )
∗ I Q′it ≤ γ′1 + δ32 LnRETIit ∗ I Q′it > γ′1 + V3t + ε3it
Journal of Cleaner Production 418 (2023) 138146
LnCO2pcit = α4i + β41 LnAFFVaddit + β42 LnInflationit + β43 LnFDIit
+ β44 LnGIit + β45 LnDEIit + β46 LnRETIxLnDEIit
( ) ( )
+ δ41 Ln RETIit ∗ I Q′it ≤ γ′1 + δ42 LnRETIit ∗ I γ′1 < Q′it ≤ γ′2
( )
+ δ43 Ln RETIit ∗ I Q′it > γ′2 + V4t + ε4it
(10)
Where the threshold variable (Qit ), can be either LnGDPpc or LnCO2pc.
As can be seen from equation (7) to equation (10), the Renewable En­
ergy Transition Index (RETI) is used as a regime-dependent variable (Rit )
in all the regressions in the framework of FEPTR.5 However, the inde­
pendent variables (Zit ) used can be slightly different depending on
whether the outcome variable is LnGDPpc or LnCO2pc. In former case,
the variables LnTNRR, LnInflation, LnFDI, LnGI, LnDEI, and the inter­
action term LnRETIxLnDEI are used as covariates while in the latter
case, all the variables maintain exactly the same except for LnTNRR,
which is replaced by LnAFFVadd.
4.7. Methods of Moments Quantile Regression (MMQR)
After the FEPTR analysis, the MMQR econometric technique is used
to measure the non-linear and heterogenous impact of the determinants
of economic growth (LnGDPpc) and environmental degradation
(LnCO2pc) across the quantile distribution. According to Alvarado et al.
(2021) and Machado and Silva (2019), the advantages of using the
MMQR are the following: 1) The MMQR is consistent and efficient in the
presence of endogenous regressors and fixed effects. Since it utilizes a
triangular structure with respect to the model parameters, it enables the
sequential calculation of the one-step Generalized Method of Moments
(GMM) estimator (Machado and Silva, 2019), 2) The MMQR allows to
estimate marginal effect of each regressor on outcome variable across
the entire quantile distribution, 3) The MMQR considers heterogeneity
(6)
The MMQR model has the following structure:
QDv (τ|Xit ) = (αi + δi q(τ)) + X′it β + Z′it γq(τ) (11)
Where QDv (τ|Xit ) denotes the conditional quantile distribution of the
dependent variable, τ denotes quantile level, X′it represents the inde­
pendent and control variables jointly, α, δ, β, γ represent the parameters
of interest, the term (αi +δi q(τ)) is the scalar coefficient which denotes
the quantile fixed effect for individual country (Machado and Silva,
2019), and finally Z′it represents a k-vector of known components of Xit
normalized to satisfy the moment conditions (Machado and Silva,
2019).
Adapting to our regression models, the MMQR can be formulated as
follows:
QLnGDPpc (τ|α1i ,ϑ1t ,X1it )= α1i +β11 LnTNRRit +β12 LnInflationit +β13 LnFDIit
(7)
+β14 LnGIit +β15 LnRETIit +β16 LnDEIit + v1t + ε1it
(12)
QLnGDPpc (τ|α2i , ϑ2t , X2it ) = α2i + β21 LnTNRRit + β22 LnInflationit + β23 LnFDIit
+ β24 LnGIit + β25 LnRETIit + β26 LnDEIit
+ β27 (LnRETI ∗ LnDEI)it + v2t + ε2it
(8)
(13)
When the LnCO2pc is considered as an outcome variable:
(9)
5
The reason why we do not use the interaction term LnRETIxLnDEI as a
regime-dependent variable is that when it is used as such, we cannot find any
evidence of the threshold effect in any of the cases.
14
Y.K. Hwang
QLnCO2pc (τ|α3i , ϑ3t , X3it ) = α3i + β31 LnAFFVaddit + β32 LnInflationit
+ β33 LnFDIit + β34 LnGIit + β35 LnRETIit
+ β36 LnDEIit + v3t + ε3it
QLnCO2pc (τ|α4i , ϑ4t , X4it ) = α4i + β41 LnAFFVaddit + β42 LnInflationit
+ β43 LnFDIit + β44 LnGIit + β45 LnRETIit
+ β46 LnDEIit + β47 (LnRETIxLnDEI)it + v4t + ε4it
Where QLnGDPpc (τ|Xit ) and QLnCO2pc (τ|X2it ) denote the quantile distribu­
tion of LnGDPpc and LnCO2pc respectively, αki and vkt represent an
unobserved country and time fixed effect respectively, βki denotes the
coefficients to estimate, and εkit represents the error term.
By applying the MMQR, the heterogenous impacts of renewable
energy transition, the digital economy, and the synergistic effect be­
tween them can be assessed across the different quantiles of country’s
per capita GDP and per capita CO2 emissions. In other words, the MMQR
enables to examine whether the impacts of the RETI, DEI, and RETIxDEI
differ across the conditional distribution of per capita income or per
capita carbon emission level. In the MMQR analysis, the effects of
renewable energy transition and other key independent variables on
conditional distribution of dependent variable are evaluated at percen­
tiles of 0.20, 0.40, 0.60, 0.80 and 0.90 respectively.
5. Empirical results
5.1. Preliminary tests
5.1.1. Correlation matrix and VIF test
From the results of the correlation matrix (see Table A.1 in Appen­
dix), we cannot find any evidence of the collinearity issue among the
independent variables as all of they have a coefficient of correlation
below 0.7. After analyzing the correlation matrix, the Variance Inflation
Factor (VIF) test are carried out to examine the multicollinearity issue
among the variables. The results of the test indicate that there are no
multicollinearity problems (see Table A.2 in Appendix) as the individual
VIF as well as the mean value of VIF are less than 10 in which they are
generally accepted benchmark (Fuinhas et al., 2021).
5.1.2. Cross-sectional dependence (CSD) test
To check the presence of CSD between the variables in our panel
data, Pesaran Cross Sectional Dependence (CD-test) Test is used
(Pesaran et al., 2006). The results indicate that all the variables, except
the interaction term LnRETIxLnDEI, present the cross-sectional depen­
dence at the 1% significance level, rejecting the null hypothesis of
cross-sectional independence (see Table A.3 in Appendix), meaning that
all the LAC countries in this study share some common shocks or char­
acteristics that make them dependent on each other.
5.1.3. Panel unit root test
After confirming the existence of the CSD in panel data, the statio­
narity of the variables is examined. For this purpose, both the first-
(Maddala and Wu, 1999) and second-generation panel unit root test
(CIPS) developed by Pesaran (2007) are conducted with two different
specifications: with and without trend (see Table A.4 in Appendix). The
first-generation Maddala and Wu Panel Unit Root test shows that,
without trend, all the variables are stationary at level, but with trend,
LnGDPpc, LnCO2pc, and LnTNRR are found to have a unit root. How­
ever, after the first difference, they all become stationary at the 1%
significance level regardless of the specification. On the other hand, the
Journal of Cleaner Production 418 (2023) 138146
CIPS test indicates that the variables LnGDPpc, LNCO2pc, and LnTNRR
(14) are not stationary irrespective of considerations on trend.6 With trend,
the variable LnAFFVadd is stationary at the 5% significance level but
shows unit root problem without trend. However, after the first differ­
ence, all of them become stationary at the 1% significance level, so, the
tests indicate that the data consists of variables with I (0) and I (1) and
the unit root problem disappears when they are transformed into first
difference.
(15)
5.1.4. Panel cointegration test
After the unit root test, the next step accompanies verifying the ev­
idence of cointegration among the variables. This is a crucial step before
implementing the non-linear regression models, such as the FEPTR and
MMQR, because we cannot use the variables at first difference to get rid
of the unit root problem. However, the use of non-stationary variables
might pose a serious problem since it might produce a spurious regres­
sion, and we should deal with it before carrying out the estimation. As an
alternative to this issue, the cointegration test is performed to see
whether the variables in regression model jointly show the evidence of a
long-run equilibrium and stable relationship or not. In other words, the
cointegration test allows to examine whether a combination of non-
stationary variables is stationary or not and jointly shows a stable
pattern in the long run. If the evidence of a cointegration relationship
between the variables with a unit root (non-stationary at levels) is
confirmed, the estimation can proceed without generating spurious
estimation results. Thus, we proceed to check the cointegration among
the variables in regression models by using two different types of panel
cointegration tests: Pedroni (first-generation cointegration test) and
Westerlund cointegration tests (second-generation cointegration test).
The former does not consider the cross-sectional dependence (CSD)
while the latter takes into account CSD and is robust in the presence of
CSD. In all the regression models (see equations (1)–(4)), the results
strongly support the evidence of cointegration relationships among the
variables, as second-generation Westerlund cointegration test rejects the
null hypothesis of no cointegration at 1% significance level against the
alternative of cointegration of all panels7 (see Table A.5 in Appendix).
5.1.5. Hausman test
To verify the existence of fixed or random effects in our regression
models, Hausman test is performed. According to Hausman test, the null
hypothesis of not systematic difference in coefficients (random-effect
regression model is more appropriate) is rejected at the 1% significance
level in all the cases. Therefore, the fixed effects should be incorporated
in regression models (see Table A.6 in Appendix).
5.2. Standard linear regression models (POLS, RE, Driscoll-Kraay FE,FE-
2SLS)
The standard linear regression models, namely Pooled Ordinary
Least Squares (POLS), Random Effects (RE), Driscoll-Kraay Fixed Effects
(D-K FE), and Fixed Effects Two Stage Least Squares (FE-2SLS) estima­
tors are performed before the non-linear ones, namely MMQR and
FEPTR for comparison purposes. Prior to estimation of linear regression
models, several tests are conducted. These include Breusch and Pagan
multiplier test, Wooldridge autocorrelation test, and Modified Wald test.
6
Among the variables which are shown to be stationary at level, only LnGI
and LnAFFVadd are weakly stationary at the 10% and 5% significance level
according to the CIPS test.
7
The time trend is included in all the cointegration tests. Furthermore, the
option “all panels” is included after the command xtcointest westerlund in Stata
to establish the alternative hypothesis of cointegration of all panels because
when this option is not included, the Westerlund cointegration test evaluates
the null hypothesis of no cointegration versus an alternative one that some
panels are cointegrated.
15
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

The purpose of the tests is to determine whether static panel is preferred Table 8
over Pooled OLS (POLS), the presence of autocorrelation of order 1, and FEPTR models regression results with Rit = Ln RETIit .
heteroskedasticity respectively (Labra and Torrecillas, 2014). The re­ Yit = LnGDPpcit
sults of the tests indicate that the panel static model is preferred over
Variables Qit = LnGDPpcit Qit = LnCO2pcit
POLS and there is the existence of autocorrelation of order 1, and
groupwise heteroskedasticity in our panel data, since the null hypoth­ LnTNRR 0.039*** 0.039*** 0.018* 0.017*
(0.010) (0.010) (0.010) (0.010)
eses are rejected at the 1% significance level in all the tests (see LnInflation − 0.020** − 0.020** − 0.006 − 0.005
Table A.7-Table A.9 in Appendix). After confirmation on the existence of (0.010) (0.010) (0.009) (0.010)
heteroskedasticity, autocorrelation of order 1, cross-sectional depen­ LnFDI 0.009 (0.007) 0.009 (0.007) − 0.000 0.000
dence, and fixed effects in previous tests, the D-K estimator with FE is (0.007) (0.007)
LnGI 0.395*** 0.396*** 0.280* 0.267*
preferred as our main econometric technique to estimate the coefficients
(0.148) (0.149) (0.146) (0.147)
in linear regression models since it is robust in the presence of auto­ LnDEI − 0.028 − 0.028 0.005 0.005
correlation, heteroskedasticity, and cross-sectional dependence. Addi­ (0.031) (0.031) (0.031) (0.031)
tionally, FE-2SLS was performed to address potential endogeneity issues LnRETIxLnDEI 0.002 (0.091) − 0.093
in linear regression models by using instrumental variables. In this re­ (0.089)
cat#c.LnRETI
gard, the lagged terms of the endogenous variables LnRETI and LnDEI 0 0.000 (0.029) 0.000 (0.029) 0.132*** 0.130***
were employed as instrumental variables, as they satisfied with the (0.036) (0.036)
relevance and exogeneity conditions (the lagged terms of LnRETI and 1 ¡0.086*** ¡0.086*** 0.038 0.035
LnDEI are correlated with LnRETI and LnDEI respectively, as they (0.031) (0.032) (0.029) (0.029)
2
represent their past values, and they only affect the outcome variables ¡0.245*** ¡0.245***
(0.036) (0.036)
through LnRETI and LnDEI respectively since the lagged terms of Constant 6.780*** 6.779*** 7.343*** 7.395***
LnRETI and LnDEI are uncorrelated with the error term). (0.608) (0.611) (0.602) (0.604)
Table 7 shows that renewable energy transition (LnRETI) has a sig­ 2
R (within) 0.837 0.837 0.836 0.836
nificant positive impact on economic growth (LnGDPpc) only in POLS. Number of 306 306 306 306
Obs.
In other static panel regressions remaining, its impact is not statistically
significant. This finding is somewhat similar to the previous study of Yit = LnCO2pcit
Charfeddine and Kahia (2019), who found that renewable energy con­ Variables Qit = LnGDPpcit Qit = LnCO2pcit
sumption had only a weak impact on the economic growth in LnAFFVadd − 0.240*** − 0.238*** − 0.144*** − 0.147***
resource-rich middle-income countries such as 24 MENA nations. On the (0.047) (0.047) (0.043) (0.043)
other hand, renewable energy transition is found to have a negative LnInflation − 0.005 − 0.005 − 0.018 − 0.016
effect on CO2 emissions in the LAC region at the significance level of at (0.016) (0.016) (0.014) (0.014)
LnFDI 0.031*** 0.031*** 0.000 (0.011) 0.000 (0.011)
least 5% level in all standard linear regression models. Unlike the nexus
(0.012) (0.012)
between energy transition and economic growth, there is a broad LnGI − 0.065 − 0.055 − 0.050 − 0.076
consensus among the researchers about the beneficial impact of (0.236) (0.238) (0.206) (0.207)
renewable energy sources on improving environmental quality and LnDEI − 0.008 − 0.009 0.001 (0.045) 0.001 (0.045)
(0.052) (0.052)
reducing air pollution levels. The study of Dong et al. (2018) demon­
LnRETIxLnDEI 0.069 − 0.147
strated that renewable energy consumption contributed significantly to (0.152) (0.130)
reducing carbon emissions in 128 countries in the globe. Mohsin et al. cat#c.LnRETI
(2021) reached a similar conclusion that renewable energy consumption 0 0.046 0.046 0.142*** 0.138***
led to a significant reduction of GHG emissions in 25 Asian countries. (0.069) (0.069) (0.050) (0.050)
1
As for the digital economy, its positive effect on economic growth is
¡0.102** ¡0.101** ¡0.103** ¡0.107**
(0.049) (0.049) (0.042) (0.042)
found in all linear regression models, although the significance level of 2 ¡0.575*** ¡0.578***
LnDEI is somewhat lower in Driscoll-Kraay FE estimator (statistically (0.064) (0.064)
significant at the 10% level). With regard to the impact of the digital Constant 1.014 0.969 0.906 (0.868) 1.019 (0.874)
(0.992) (0.998)
economy on CO2 emissions, the positive and significant impact at 10% 2
R (within) 0.503 0.503 0.610 0.612
level is found only in POLS while in other panel static models, the co­ Number of 306 306 306 306
efficient of LnDEI is not statistically significant. Regarding the syner­ Obs.
gistic effect between renewable energy transition and the digital
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively
economy on economic growth and environmental sustainability, this
and time dummies were included during the estimation; Yit denotes the
impact is not observed in any of the estimations, except for FE-2SLS. In dependent variable. The values in parentheses indicate the standard error.
FE-2SLS estimation, the elasticity of LnRETIxLnDEI is found to be
negative and statistically significant at the 5% level in both cases,
transition and economic growth or environmental degradation shows
indicating that the synergistic effect has a negative impact on economic
non-linearity (see Appendix).
growth but a positive impact on environmental sustainability. In sum,
When considering per capita GDP as the dependent and threshold
the results of standard linear regressions indicate that the positive im­
variable (with Yit = Ln GDPpcit , Qit = Ln GDPpcit ), a double-threshold
pacts of the transition to renewable energy on enhancing environmental
effect is found at the 10% significance level (see Table A.11 and A.12
quality and the digital economy on promoting economic growth are
in Appendix). Table 8 shows that the impact of renewable energy
observed consistently when considering the LAC region as a whole.
transition on economic growth is not statistically significant when per
However, when it comes to the synergistic effect, its significant impact is
capita income is below the first threshold (LnGDPpc <8.530). However,
only observed in FE-2SLS estimation.
when LnGDPpc is in between the first and second threshold (8.530
≤LnGDPpc <9.362), the impact of renewable energy transition becomes
5.3. Fixed-Effect Panel Threshold Regression (FEPTR)
significant and negatively affects economic growth with a coefficient of
− 0.086. When LnGDPpc crosses the second threshold value (9.362), the
5.3.1. FEPTR estimation results without interaction term (LnRETIxLnDEI)
coefficient of LnRETI becomes even more negative (− 0.245) and sta­
Before conducting the FEPTR, a threshold effect test is performed to
tistically significant at the 1% significance level.
determine whether the relationship between renewable energy

16
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

When considering per capita GDP and per capita CO2 emissions as
the dependent and the threshold variable respectively (Yit = Ln GDPpcit ,
Qit = Ln CO2pcit ), a single-threshold model is found to be appropriate at
the 5% significance level based on the threshold effect test (see
Table A.17 and A.18 in Appendix). Table 8 shows that the impact of
renewable energy transition on economic growth is positive and statis­
tically significant when LnCO2 is less than 0.545, but as LnCO2pc ex­
ceeds the threshold value, its positive impact becomes statistically not
significant. Specifically, when LnCO2pc is below the threshold (1.076),
the coefficient of LnRETI is 0.132 but when LnCO2pc is above the
threshold, the positive effect of renewable energy transition on eco­
nomic growth disappears.
When considering per capita CO2 emissions and per capita GDP as
the dependent and threshold variable respectively (Yit = Ln CO2pcit ,
Qit = Ln GDPpcit ), a single-threshold effect is found at the 1% signifi­
cance level (see Table A.23 and A.24 in Appendix). According to the
FEPTR estimates shown in Table 8, the impact of LnRETI on LnCO2pc is
not significant when LnGDPpc is less than 8.516. However, when
LnGDPpc exceeds the threshold value, the coefficient of LnRETI turns
negative and statistically significant at the 5% level (− 0.102).
When considering per capita CO2 emissions as the dependent and
threshold variable (Yit = Ln CO2pcit , Qit = Ln CO2pcit ), a double-
threshold effect is found at the 1% significance level (see Table A.29
and A.30 in Appendix). According to the FETHR estimates in Table 8,
when LnCO2pc is below the first threshold (0.123), the coefficient of
LnRETI is 0.142, indicating that renewable energy transition leads to
environmental deterioration. However, when LnCO2pc is between
0.123 and 1.153, the coefficient of LnRETI turns negative and is statis­
tically significant at the 5% level (− 0.103), which means that renewable
energy transition contributes to improving environmental sustainability
in the LAC region within this range of CO2 values. When LnCO2pc ex­
ceeds the second threshold (1.153), the coefficient of LnRETI becomes
even more negative than the previous one and statistically significant at
the 1% significance level (− 0.575), thus showing the gradual increasing
effect of renewable energy transition on environmental and air quality
improvement in terms of CO2 emissions reduction when countries’
emission level is already high.
Lastly, when interaction term, namely LnRETIxLnDEI is included in
regression models, the estimates of LnRETI obtained are practically the
same as previous FEPTR analysis without LnRETIxLnDEI as shown in
Table 8, confirming the robustness of our empirical findings.
5.3.2. FEPTR estimation results with the regime-dependent variable
replaced
To further ensure the robustness and reliability of our empirical
findings, additional FEPTR analysis is conducted by replacing LnRETI
with LnREShareTFEC (which denotes the renewable energy share in
total final energy consumption) as a regime-dependent variable. In the
previous studies, important authors have used the share of renewable
energy consumption in the energy mix as a proxy for the clean energy
transition in their seminal works (Doytch and Narayan, 2021; Inglesi-­
Lotz, 2016). Therefore, LnREShareTFEC can be a good proxy for LnRETI.
Table 9 shows that even with the change of regime-dependent variable
in regression models, the main estimation results remain stable. LnRE­
ShareTFEC, like LnRETI, shows an important threshold effect on eco­
nomic growth in terms of per capita income and per capita CO2 emission
levels. The negative impact of LnREShareTFEC on economic growth is
observed, and its elasticity decreases as a country’s LnGDPpc or
LnCO2pc increases. The only difference is that when LnCO2pc is
considered as a threshold variable, no threshold effect is detected or the
negative impacts of LnREShareTFEC on economic growth gradually at­
tenuates. On the other hand, LnREShareTFEC negatively affects CO2
emissions and shows an important threshold effect depending on a
country’s income or emission levels, which is consistent with the pre­
vious estimation results when using LnRETI as a regime-dependent
variable. The only difference is that the beneficial impact of

Table 9
FEPTR model regression results with Rit = Ln REShareTFECit .
Yit = LnGDPpcit

Variables Qit = LnGDPpcit Qit = LnCO2pcit

LnTNRR 0.038*** (0.009) 0.038*** (0.009) 0.024** (0.010) 0.024** (0.010)

LnInflation − 0.006 (0.008) − 0.005 (0.008) − 0.005 (0.010) − 0.005 (0.010)
LnFDI − 0.001 (0.006) − 0.000 (0.006) − 0.005 (0.007) − 0.004 (0.007)
LnGI 0.428*** (0.125) 0.446*** (0.125) 0.640*** (0.147) 0.653*** (0.148)
LnDEI 0.016 (0.026) 0.028 (0.028) 0.019 (0.031) 0.029 (0.033)
LnREShareTFECxLnDEI − 0.053 (0.040) − 0.045 (0.047)
cat#c.LnREShareTFEC
0 ¡0.207*** (0.021) ¡0.207*** (0.021) No threshold effect ¡0.136*** (0.025)
1 ¡0.173*** (0.021) ¡0.174*** (0.021) ¡0.103*** (0.026)
2 ¡0.118*** (0.021) ¡0.118*** (0.021) ¡0.059** (0.029)
Constant 7.346*** (0.500) 7.283*** 6.263*** (0.601) 6.214*** (0.604)
R2 (within) 0.883 0.884 0.838 0.838
Number of Obs. 306 306 306 306

Yit = LnCO2pcit

Variables Qit = LnGDPpcit Qit = LnCO2pcit

LnAFFVadd − 0.095** (0.039) − 0.100** (0.039) − 0.198*** (0.036) − 0.170*** (0.036)

LnInflation 0.011 (0.012) 0.011 (0.012) 0.011 (0.012) 0.013 (0.012)
LnFDI 0.021** (0.009) 0.022** (0.009) 0.014 (0.009) 0.015* (0.009)
LnGI − 0.229 (0.181) − 0.239 (0.181) − 0.070 (0.171) − 0.048 (0.171)
LnDEI − 0.008 (0.040) − 0.021 (0.042) 0.011 (0.039) − 0.006 (0.041)
LnREShareTFECxLnDEI 0.050 (0.061) − 0.084 (0.060)
cat#c.LnREShareTFEC
0 ¡0.444*** (0.032) ¡0.444*** (0.032) ¡0.516*** (0.031) ¡0.532*** (0.032)
1 ¡0.390*** (0.033) ¡0.389*** (0.033) ¡0.465*** (0.030) ¡0.465*** (0.030)
2 ¡0.421*** (0.031) ¡0.420*** (0.031)
Constant 2.920*** (0.748) 2.960*** (0.746) 2.659*** (0.710) 2.474*** (0.710)
R2 (within) 0.698 0.702 0.713 0.715
Number of Obs. 306 306 306 306

Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation; Yit denotes the dependent
variable. The values in parentheses indicate the standard error.

17
Y.K. Hwang
LnREShareTFEC decreases when LnGDPpc or LnCO2pc increases.
However, the sign and the significance level do not differ significantly
when the regime-dependent variable changes, confirming the robust­
ness of the empirical findings.
5.4. Method of Moments Quantile Regression (MMQR)
5.4.1. MMQR without time fixed effect
The standard linear regression models may not properly account for
the heterogenous distribution of panel data and can lead to biased
estimation results when outliers are present or when the elasticity of the
variable of interest differ significantly across different points of the
outcome variable (Alvarado et al., 2021). To address this issue, the
MMQR is used in our analysis for estimating the asymmetric and
non-linear effect of renewable energy transition and other covariates on
economic growth and CO2 emissions across the quantile distribution.
Before estimating the MMQR, a battery of tests, including Swamey’s test
(Pesaran and Yamagata, 2008), Shapiro-Wilk W test (Royston, 1983),
Skewness-Kurtosis test (D’agostino et al., 1990; Royston, 1992) are
carried out to evaluate slope homogeneity and the normal distribution of
the dependent variable (see Table A.59-A.61 and Figure A.1 in Appen­
dix). The results confirm slope heterogeneity and non-normality of
LnGDPpc and LnCO2pc as the null hypotheses of Swamey, Shapiro-Wilk
and Skewness-Kurtosis tests are rejected at the 1% significance level
respectively, satisfying prerequisites for the application of the MMQR to
our analysis objective (Fuinhas and Marques, 2019). Table 10 shows the
estimation results of the MMQR without time fixed effects. The MMQR is
conducted at the 20th,40th,60th, 80th, and 90th percentiles of the
conditional distribution. The results indicate a significant and positive
impact of renewable energy transition on economic growth across all
quantiles, but this promotional effect of LnRETI is especially stronger at
lower quantiles. However, as one moves along the quantile distribution,
the positive effect of renewable energy transition on economic growth
gradually decreases. These findings highlight the fact that in the LAC
region, the beneficial impacts of renewable energy transition on eco­
nomic growth in lower-income countries are much larger than
high-income countries. The possible explanation is that the integration
of renewable energy sources in the electricity mix can greatly contribute
to improving energy access and reducing energy poverty for a large
share of vulnerable populations living in low-income countries through
off-grid and distributed electricity generation, which might eventually
lead to improving standard of livings of poor households and local
economic development (Urban, 2014). The evidence of the beneficial
impact of renewable energy-based off-grid electricity generation on
economic growth and poverty alleviation can be found in the study
conducted by Wirawan and Gultom (2021). The researchers investigated
the potential effects of renewable energy-based rural electrification on
poverty alleviation in 217 remote non-grid villages in Indonesia. The
findings of their study indicate that renewable-based off-grid electrifi­
cation has significant positive effects. Specifically, it reduces the poverty
level, decreases the number of individuals relying on health insurance,
and contributes to an increase in the number of small industries, which
ultimately has a positive impact on the local economy. Furthermore, the
potential marginal benefits in low-income countries that can be obtained
from renewable energy transition at its early stage of development is
much larger than high-income countries, this is because the lack of
infrastructure and low-quality electricity networks predominant in
low-income countries can be greatly reduced or enhanced by the inte­
gration of renewables in the electricity mix (Vanegas Cantarero, 2020).
But for high-income countries in the LAC region, the relatively small
impact of renewable energy transition on economic growth can be
explained by the fact that in these countries, non-renewable and
polluting fossil fuels are the major energy sources responsible for their
energy needs, to carry out economic activities in a cost-effective way and
to ensure economic growth in the short-term, which prevents the suc­
cessful implementation of renewable energy transition requiring
18
Journal of Cleaner Production 418 (2023) 138146
long-term period (Henderson and Sen, 2021). Also in high-income
countries, the existing energy system and infrastructures are already
highly adapted to fossil fuel energy sources (large thermal power plants,
large gas pipelines in Argentina and Chile), and in case these economies
decide to phase out fossil-fuels and move to renewable energy transition,
a significant part of these assets has a high risk of becoming stranded
(ECLAC, 2022). For this reason, relatively wealthy countries in the LAC
region may decide to postpone renewable energy transition and remain
locked in legacy technologies (fossil fuel-based technologies). Regarding
the impact of the digital economy on economic growth, this study finds a
positive and statistically significant effect of LnDEI at the 1% signifi­
cance level on LnGDPpc in all quantiles. However, this effect decreases
as the quantile increases. These findings demonstrated that the process
of digitalization contributes to accelerating economic growth in all
countries of the Latin American and the Caribbean region, but its impact
is particularly strong for the countries belonging to the lower quantile.
This is because the marginal benefits of the digital economy in
low-income countries are greater than high-income countries due to the
lower degree of digitalization. Therefore, the potential gains from
incorporating digital technologies into their workspace and production
processes might be enormous in terms of improving labor productivity,
thus resulting in economic growth for the country. It is worth noting that
the coefficient of LnDEI is systematically larger than that of LnRETI in
MMQR1 and MMQR2 as shown in Table 10, which means that the digital
economy has a much larger impact on promoting economic growth than
renewable energy transition in the LAC region.
Regarding the effect of the interaction term (LnRETIxLnDEI) on
economic growth, we can only verify its significant impact at the 20th
quantile (1.815 at the 5% significance level). This finding indicates that
in low-income LAC countries, the synergistic effect between the digital
economy and renewable energy transition on accelerating economic
growth is especially high. On the other hand, the finding also suggests
that most LAC countries are still far from taking full advantages of the
synergy effect between the digital economy and renewable energy
transition to accelerate economic growth as the coefficient of LnRE­
TIxLnDEI is statistically not significant at other quantile groups. With
respect to the control variables, we found no significant impact of total
natural resources rents on economic growth across all quantiles. The
finding is somewhat unexpected since the fiscal revenues derived from
the exploitation of natural resources occupy a significant portion of
governments’ budgets in many LAC countries. The possible explanation
is that heavy reliance on revenues obtained from exploitation of natural
resources in their fiscal capacity, such as production of oil and natural
gas and extraction of minerals, makes LAC countries highly vulnerable
and susceptible to changes in global commodity market prices, which
worsens the macroeconomic stability of these nations and acts as a
stumbling block to the successful diversification of their economy. With
regard to LnInflation, its impact on economic growth is positive and
statistically significant at the 1% significance level across all quantiles.
The positive effect of inflation on economic growth in the LAC region is
somewhat surprising, given that a high price level is generally expected
to have a detrimental impact on economic growth. This is due to the
reduction in the real purchasing power of consumers, which can result in
decreased consumption and lower aggregate demand. Moreover, high
inflation can generate uncertainty among investors, potentially leading
to reduced savings and investments, thereby further dampening eco­
nomic growth. The possible explanation for the positive effect of infla­
tion on economic growth is that many LAC countries are net exporters of
commodities, and a high level of inflation, to some extent, might benefit
them in terms of trade since they can sell commodities like oil, gas,
agricultural goods, and minerals at a much higher price to other coun­
tries, as happened in Commodity boom period from 2000 to 2014. Also
we came to recognize that the real impact of inflation on economic
growth in the LAC region may not be fully captured due to the limited
number of varibles used in the regression model or the specific model
specifications used in this study. As for the potential of globalization, a
 Y.K. Hwang
Table 10
MMQR estimation results (without time fixed effects).
LnGDPpc

Variables Location Scale Q20 Q40 Q60 Q80 Q90

MMQR 1 MMQR 2 MMQR 1 MMQR 2 MMQR 1 MMQR 2 MMQR 1 MMQR 2 MMQR 1 MMQR 2 MMQR 1 MMQR 2 MMQR 1 MMQR 2

LnTNRR − 0.013 − 0.010 − 0.006 − 0.007 − 0.006 − 0.001 − 0.012 − 0.008 − 0.016 − 0.013 − 0.020 − 0.018 − 0.022 − 0.020
(0.020) (0.021) (0.011) (0.012) (0.028) (0.031) (0.021) (0.023) (0.019) (0.019 (0.019) (0.019) (0.020) (0.020)
LnInflation 0.206*** 0.204*** 0.004*** 0.005 0.202*** 0.199*** 0.205*** 0.203*** 0.207*** 0.207*** 0.210*** 0.210*** 0.211*** 0.211***
(0.052) (0.052) (0.029) (0.031) (0.073) (0.076) (0.055) (0.057) (0.048) (0.047) (0.050) (0.047) (0.052) (0.050)
LnFDI − 0.102 − 0.104 − 0.042 − 0.040 − 0.052 − 0.056 − 0.094 − 0.093 − 0.123* − 0.123* − 0.151** − 0.150** − 0.165** − 0.162**
(0.074) (0.075) (0.040) (0.044) (0.103) (0.109) (0.077) (0.081) (0.068) (0.068) (0.070) (0.068) (0.074) (0.071)
LnGI 4.012*** 4.070*** − 0.612*** − 0.615*** 4.743*** 4.800*** 4.134*** 4.237*** 3.712*** 3.786*** 3.301*** 3.364*** 3.112*** 3.190***
(0.337) (0.345) (0.185) (0.202) (0.471) (0.505) (0.362) (0.383) (0.315) (0.315) (0.322) (0.315) (0.336) (0.326)
LnRETI 0.336*** 0.334*** ¡0.078*** ¡0.073 0.429*** 0.421*** 0.351*** 0.354*** 0.297*** 0.300*** 0.245*** 0.249*** 0.220** 0.229**
(0.092) (0.095) (0.050) (0.056) (0.128) (0.139) (0.097) (0.104) (0.085) (0.086) (0.087) (0.086) (0.092) (0.091)
LnDEI 0.800*** 0.796*** ¡0.199*** ¡0.166 1.037*** 0.993*** 0.839*** 0.841*** 0.702*** 0.720*** 0.568*** 0.606*** 0.507** 0.559***
(0.197) (0.193) (0.108) (0.113) (0.275) (0.283) (0.209) (0.211) (0.184) (0.175) (0.188) (0.175) (0.198) (0.185)
LnRETIxLnDEI 0.742 ¡0.905*** 1.815** 0.988 0.324 ¡0.297 ¡0.552
(0.617) (0.361) (0.902) (0.679) (0.561) (0.561) (0.585)
Constant − 7.395*** − 7.644*** 2.734*** 2.781*** − 10.663*** − 10.945 − 7.942*** − 8.402*** − 6.504*** − 6.362*** − 4.215*** − 4.452 − 3.372** − 3.666***
(1.454) (1.493) (0.798) (0.874) (2.030) (1.562) (1.658) (1.359) (1.360) (1.390) (1.450) (1.410)
Number of Obs. 306 306 306 306 306 306 306 306 306 306 306 306 306 306
19

LnCO2pc

Variables Location Scale Q20 Q40 Q60 Q80 Q90

MMQR 3 MMQR 4 MMQR 3 MMQR 4 MMQR 3 MMQR 4 MMQR 3 MMQR 4 MMQR 3 MMQR 4 MMQR 3 MMQR 4 MMQR 3 MMQR 4

LnAFFVadd − 0.974*** − 0.976*** − 0.054* − 0.047 − 0.916*** − 0.927*** − 0.956*** − 0.961*** − 0.991*** − 0.989*** − 1.031*** − 1.023*** − 1.056*** − 1.051***
(0.051) (0.051) (0.031) (0.031) (0.057) (0.057) (0.051) (0.050) (0.053) (0.053) (0.064) (0.063) (0.075) (0.076)
LnInflation 0.174*** 0.175*** − 0.077** − 0.086** 0.256*** 0.264*** 0.199*** 0.201*** 0.149*** 0.151* 0.093 0.087 0.056 0.036
(0.054) (0.057) (0.033) (0.035) (0.061) (0.065) (0.054) (0.057) (0.057) (0.060) (0.069) (0.072) (0.080) (0.086)
LnFDI 0.003 0.006 − 0.010 − 0.012 0.014 0.018 0.007 0.009 − 0.000 0.002 − 0.008 − 0.006 − 0.013 − 0.013
(0.033) (0.034) (0.020) (0.021) (0.037) (0.038) (0.032) (0.033) (0.034) (0.035) (0.041) (0.042) (0.048) (0.050)
LnGI 1.599*** 1.542*** 0.078 0.064 1.516*** 1.477*** 1.573*** 1.523*** 1.625*** 1.560*** 1.682*** 1.608*** 1.720*** 1.645***
(0.246) (0.252) (0.151) (0.155) (0.275) (0.282) (0.244) (0.251) (0.257) (0.261) (0.311) (0.315) (0.360) (0.378)
LnRETI ¡0.638*** ¡0.636*** ¡0.044 ¡0.037 ¡0.591*** ¡0.598*** ¡0.624*** ¡0.625*** ¡0.653*** ¡0.647*** ¡0.685*** ¡0.674*** ¡0.707*** ¡0.696***
(0.069) (0.070) (0.043) (0.043) (0.078) (0.078) (0.069) (0.069) (0.073) (0.072) (0.088) (0.087) (0.102) (0.105)

Journal of Cleaner Production 418 (2023) 138146

LnDEI 0.228* 0.230* ¡0.015 0.026 0.243* 0.203 0.233* 0.222* 0.223* 0.237* 0.212 0.257 0.205 0.273
(0.124) (0.126) (0.076) (0.077) (0.138) (0.141) (0.123) (0.125) (0.129) (0.131) (0.156) (0.157) (0.181) (0.189)
LnRETIxLnDEI ¡0.651* 0.083 ¡0.737* ¡0.676* ¡0.627 ¡0.565 ¡0.515
(0.380) (0.234) (0.425) (0.378) (0.394) (0.475) (0.570)
Constant − 5.130*** − 4.887*** 0.148 0.256 − 5.288*** − 5.151*** − 5.179*** − 4.965*** − 5.081*** − 4.816*** − 4.793*** − 4.624*** − 4.902*** − 4.472***
(1.081) (1.108) (0.663) (0.681) (1.209) (1.236) (1.074) (1.101) (1.130) (1.147) (1.365) (1.382) (1.582) (1.659)
Number of Obs. 306 306 306 306 306 306 306 306 306 306 306 306 306

Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation; MMQR 1- MMQR 4 correspond to equation 12–15 without time fixed effects in section
4.6 respectively. The values in parentheses indicate the standard error.
Y.K. Hwang
very strong positive and statistically significant effect at the 1% level on
economic growth is found in all quantiles. The results are consistent with
the previous findings of Koengkan and Fuinhas (2022), who confirmed
that globalization had a positive impact on the improvement of the in­
come level in LAC countries. Regarding FDI, we cannot find any sig­
nificant impact on economic growth across the quantile distribution.
When considering LnCO2pc as a dependent variable, a negative and
statistically significant effect of LnRETI at the 1% significance level is
found across all quantiles, and its negative impact on CO2 emissions
gradually increases as the quantile increases. The finding indicates that
the positive contribution of renewable energy transition to improving
environmental sustainability in the LAC region is very robust, and its
impact on improving environmental quality in countries with high per
capita CO2 emissions is particularly stronger than those with low per
capita CO2 emissions. When it comes to the digital economy, a positive
and statistically significant impact at the 10% significance level is found
at the 40th and 60th quantiles but no meaningful effect is found in other
quantile groups. The finding shows that the potential benefits derived
from the digital economy in terms of energy and resource use efficiency
improvement (thus less energy demand) are not sufficiently exploited in
the LAC region and the negative impact of the digital economy on the
environment is especially noticeable in middle-emission countries.
Concerning the synergistic effect between the digital economy and
renewable energy transition, LnRETIxLnDEI is found to have a negative
and statistically significant impact at the 10% significance level on
LnCO2pc at the 20th and 40th quantiles. The finding indicates that
renewable energy transition if combined with the digital economy,
might lead to a large reduction in carbon dioxide emissions in the region,
which is especially true for economies with a moderate level of CO2
emissions in the LAC region. With respect to the control variables, the
results indicate that globalization leads to significantly increase carbon
dioxide emissions in all quantiles, and this effect increases as ones move
along the quantile distribution. Meanwhile, agriculture, forestry, and
fishing activities significantly reduce CO2 emissions at all quantile
groups. These findings suggest that globalization is not effective in
improving environmental quality in the LAC region through the import
of clean energy technologies or attracting environmentally friendly
firms. The negative coefficient of LnAFFVadd indicates that the activ­
ities related to the production of crops, livestock, and fishing are rela­
tively environmentally friendly in the LAC region.
Regarding inflation, a positive and statistically significant effect on
carbon dioxide emissions is found at the 20th, 40th, and 60th quantiles.
The possible explanation can be attributed to the fuel stacking problem,
which implies the simultaneous use of traditional biomass energy
sources with modern energy sources (electricity) in many households in
the LAC region. As the energy price increases, many households opt for
cheaper energy sources such as firewood and organic wastes (which is
highly polluting) to cooking and heating rather than expensive modern
energy sources. Lastly, no significant impact of FDI on carbon dioxide
emissions in the LAC region is found across the quantile distribution.
5.4.2. MMQR with time fixed effect
To confirm the robustness of our empirical findings in MMQR anal­
ysis, time fixed effects are included in regression models. The estimation
results of MMQR with time fixed effects shown in Table 11 indicate that
the empirical findings of this study are robust since no significant
changes are observed in the signs and elasticities of variables compared
to the previous estimation results of the MMQR without time fixed
effects.
5.5. Discussion of findings
Based on the previous estimation results, several important outcomes
can be highlighted as follows:
First, the synergistic effect between renewable energy transition and
the digital economy in the LAC region is not widely confirmed, and its
20
Journal of Cleaner Production 418 (2023) 138146
impact on economic growth and environmental sustainability is inter­
dependent in income and emission levels, thus verifying only partially
the Hypothesis 1 of this study. Linear regressions estimate that the
synergy between them is not significant to economic growth or envi­
ronmental sustainability in 18 combined LAC countries. However, the
non-linearity and heterogeneity of the synergistic effect between
renewable energy transition and the digital economy are found through
the MMQR analysis, showing that the synergistic effect on promoting
economic growth is found at the 20th quantile only (i.e., low-income
countries), while the synergistic effect on improving environmental
quality is found at the 40th, the 60th and the 80th quantiles8 (i.e., low-,
middle-, and high- CO2 emitting countries). The estimation results are
somewhat contrary to the findings in the study of Shahbaz et al. (2022),
who confirmed that the digital economy positively affected energy
transition in 72 countries, and its impact was especially strong in
high-income countries. The contradiction can be explained by the fact
that many LAC economies’ lacks high-quality energy and digital infra­
structure, as well as qualified workers with specialized knowledge and
engineering skills capable of managing advanced clean energy and
digital technologies (important skill gaps), which are stumbling blocks
to broad application of advanced digital technologies in the energy
system of many LAC countries. This in turn, impedes the rapid inte­
gration of renewable energy sources in the energy mix, as their high
unpredictability features such as intermittency and variability are not
properly addressed. Consequently, the process of renewable energy
transition is significantly delayed. Furthermore, the fiscal capacity of
many LAC governments limited in funding large-scale renewable energy
deployment and renovating extant digital infrastructure, above all
things the production process and local value chain heavily locked-in to
fossil fuel technologies, makes it difficult to create the synergy between
renewable energy transition and the digital economy in the LAC region.
The strong impact of the synergistic effect between renewable energy
transition and the digital economy on promoting economic growth
found in low-income LAC countries due to the fact that the marginal
benefits derived from additional energy access and digitalization in
these countries are significantly higher than in high-income countries in
terms of productivity gains, improvements in standards of living and
reduction of energy poverty reduction. Xu et al. (2022) also found, in
their study of 109 countries, that the impact of digitalization on
reducing energy intensity and optimizing energy system structure was
especially strong in low-income and developing countries. The finding of
synergistic effect between renewable energy transition and the digital
economy on reducing CO2 emissions in low- and middle-emission LAC
countries in this study contrasts with that of Lange et al. (2020), who
argued that digitalization had no effect on the decoupling of economic
growth from energy consumption due to the rebound effect in
high-energy- consuming and industrialized countries.
Second, this study reveals an important threshold effect of renewable
energy transition on economic growth and CO2 emissions in the LAC
region, and certainly proved the validity of Hypothesis 2. The FEPTR
analysis shows that the negative effect of renewable energy transition on
economic growth becomes larger as income levels increases. The posi­
tive impact of renewable energy transition on economic growth is
observed only when emissions levels are below the threshold, and this
positive impact disappears when emissions levels exceed the threshold.
This is because LAC countries with relatively high-income levels have
higher per capita energy consumption compared with low-income
countries. As the major sources of total final energy consumption in
high-income countries in the LAC region are hydrocarbons and given the
intermittency of renewables, the relative importance of renewable en­
ergy as its contribution to economic growth decreases. Doytch and
8
This is when the time fixed effects were considered in regression models.
However, when they were not taken into account, the synergistic effect was
found at the 20th and the 40th quantiles.
 Y.K. Hwang
Table 11
MMQR estimation results (with time fixed effects).
LnGDPpc

Variables Location Scale Q20 Q40 Q60 Q80 Q90

MMQR 1 MMQR 2 MMQR 1 MMQR 2 MMQR 1 MMQR 2 MMQR 1 MMQR 2 MMQR 1 MMQR 2 MMQR 1 MMQR 2 MMQR 1 MMQR 2

LnTNRR − 0.000 0.003 − 0.010 − 0.012 0.011 0.018 0.002 0.006 − 0.005 − 0.003 − 0.011 − 0.011 − 0.014 − 0.015
(0.021) (0.021) (0.011) (0.029) (0.030) (0.022) (0.020) (0.020) (0.020) (0.021) (0.021) (0.022)
LnInflation 0.232*** 0.233*** 0.003 0.005 0.228*** 0.227*** 0.231*** 0.232 0.234*** 0.235*** 0.236*** 0.238*** 0.237*** 0.240***
(0.060) (0.059) (0.032) (0.083) (0.081) (0.063) (0.056) (0.055) (0.058) (0.058) (0.061) (0.061)
LnFDI − 0.085 − 0.087 − 0.035 − 0.034 − 0.042 − 0.045 − 0.077 − 0.078 − 0.103 − 0.104 − 0.125* − 0.125* − 0.136* − 0.136***
(0.071) (0.072) (0.038) (0.099) (0.099) (0.075) (0.067) (0.068) (0.069) (0.071) (0.073) (0.075)
LnGI 4.134*** 4.162*** − 0.694*** − 0.718 4.987*** 5.049*** 4.294*** 4.346 3.786*** 3.801*** 3.338*** 3.354*** 3.132*** 3.124***
(0.366) (0.373) (0.197) (0.506) (0.512) (0.397) (0.346) (0.356) (0.354) (0.367) (0.369) (0.385)
LnRETI 0.307*** 0.308*** ¡0.058 ¡0.057 0.378*** 0.378*** 0.320*** 0.323 0.278*** 0.280*** 0.240*** 0.245** 0.223** 0.226**
(0.094) (0.098) (0.051) (0.131) (0.134) (0.100) (0.088) (0.092) (0.091) (0.096) (0.096) (0.102)
LnDEI 0.902*** 0.896*** ¡0.141 ¡0.098 1.075*** 1.017*** 0.934*** 0.921 0.831*** 0.847*** 0.740*** 0.785*** 0.698*** 0.754***
(0.204) (0.196) (0.110) (0.283) (0.269) (0.216) (0.191) (0.185) (0.197) (0.192) (0.207) (0.205)
LnRETIxLnDEI 0.717 ¡1.086 2.059** 0.996 0.170 ¡0.506 ¡0.854
(0.607) (0.833) (0.580) (0.598) (0.628)
Constant − 7.831*** − 7.977*** 3.094*** 3.240*** − 11.633*** − 11.980*** − 8.543*** − 8.810*** − 6.283*** − 6.347*** − 4.285*** − 4.331*** − 3.368** − 3.291**
(1.545) (1.561) (0.830) (0.848) (2.136) (2.144) (1.680) (1.698) (1.462) (1.496) (1.495) (1.540) (1.557) (1.612)
Number of Obs. 306 306 306 306 306 306 306 306 306 306 306 306 306 306
21

LnCO2pc

Variables Location Scale Q20 Q40 Q60 Q80 Q90

MMQR 3 MMQR 4 MMQR 3 MMQR 4 MMQR 3 MMQR 4 MMQR 3 MMQR 4 MMQR 3 MMQR 4 MMQR 3 MMQR 4 MMQR 3 MMQR 4

LnAFFVadd − 0.970*** − 0.972*** − 0.081** − 0.073** − 0.880*** − 0.890*** − 0.947*** − 0.955*** − 0.995*** − 0.996*** − 1.053*** − 1.052*** − 1.099*** − 1.090***
(0.051) (0.050) (0.031) (0.030) (0.064) (0.061) (0.052) (0.051) (0.052) (0.052) (0.059) (0.061) (0.070) (0.071)
LnInflation 0.165*** 0.165** − 0.089** − 0.100** 0.265*** 0.279*** 0.191*** 0.188*** 0.138** 0.132** 0.074 0.055 0.023 0.002
(0.061) (0.064) (0.037) (0.039) (0.076) (0.078) (0.062) (0.065) (0.062) (0.066) (0.071) (0.079) (0.084) (0.091)
LnFDI 0.010 0.014 − 0.014 − 0.015 0.027 0.030 0.015 0.017 0.006 0.009 − 0.004 − 0.003 − 0.013 − 0.010
(0.036) (0.036) (0.022) (0.022) (0.045) (0.044) (0.037) (0.036) (0.037) (0.037) (0.042) (0.044) (0.050) (0.051)
LnGI 1.924*** 1.885*** 0.379** 0.363** 1.502*** 1.471*** 1.817*** 1.800*** 2.042*** 2.002*** 2.313*** 2.284*** 2.529*** 2.473***
(0.284) (0.281) (0.172) (0.170) (0.350) (0.340) (0.289) (0.284) (0.287) (0.289) (0.327) (0.342) (0.386) (0.397)
LnRETI ¡0.683*** ¡0.685*** ¡0.111** ¡0.104** ¡0.560*** ¡0.566*** ¡0.652*** ¡0.660*** ¡0.718*** ¡0.718*** ¡0.797*** ¡0.799*** ¡0.860*** ¡0.853***
(0.074) (0.074) (0.045) (0.045) (0.092) (0.090) (0.076) (0.075) (0.075) (0.077) (0.086) (0.091) (0.101) (0.105)

Journal of Cleaner Production 418 (2023) 138146

LnDEI 0.267* 0.273** ¡0.017 0.019 0.286* 0.252 0.272* 0.269* 0.262* 0.280** 0.250 0.294* 0.241 0.304
(0.137) (0.138) (0.083) (0.084) (0.168) (0.166) (0.140) (0.139) (0.138) (0.141) (0.158) (0.167) (0.187) (0.196)
LnRETIxLnDEI ¡0.810* ¡0.046 ¡0.758 ¡0.799* ¡0.825* ¡0.860* ¡0.884
(0.426) (0.259) (0.513) (0.429) (0.436) (0.517) (0.604)
Constant − 6.321*** − 6.126*** − 0.972 − 0.856 − 5.239*** − 5.152*** − 6.046*** − 5.926*** − 6.624*** − 6.403*** − 7.318*** − 7.067*** − 7.871*** − 7.514***
(1.179) (1.175) (0.717) (0.713) (1.456) (1.417) (1.205) (1.184) (1.194) (1.204) (1.366) (1.427) (1.611) (1.663)
Number of Obs. 306 306 306 306 306 306 306 306 306 306 306 306 306 306

Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation; MMQR 1- MMQR 4 correspond to equation 12–15 in section 4.6 respectively. The
values in parentheses indicate the standard error.
Y.K. Hwang
Narayan (2021) also argued that renewable energy transition was only
effective in promoting high-growth sectors of an economy, which cor­
responds to the agriculture and mining (extractive) sectors of most LAC
nations. However, these two sectors are resource- and energy-intensive
sectors with a high reliance on hydrocarbon energy sources, such as oil
and gas, which impedes the use of renewable energy sources and po­
tential gains that can be reaped from renewable energy transition, such
as low-electricity generating costs and new jobs creation in clean energy
sector. The positive and significant impact of renewable energy transi­
tion on economic growth in countries with low levels of CO2 emissions
(thus lower per capita energy consumption due to either high energy
efficiency or low access to energy) can be explained by the fact that in
both cases, the transition to renewable energy can boost economic
growth in terms of improving further energy efficiency or offering
affordable energy services to poor segments of population of an econ­
omy. Regarding the effect of renewable energy transition on improving
environmental sustainability, the estimation results of FEPTR also
indicate a significant threshold effect. When a country’s income level is
below the threshold, no significant impact of renewable energy transi­
tion is found. However, when a country’s per capita GDP exceeds the
threshold value, the impact of renewable energy transition on CO2
emissions becomes significant, leading to an improvement in environ­
mental quality by reducing CO2 emissions. This finding is in accord with
that of Li et al. (2022. b), who found a non-linear and significant
threshold effect of renewable energy on the ecological footprint in 120
countries. They verified a positive impact of renewable energy on
environmental sustainability and emphasized that renewable energy
only became effective in improving environmental quality when a
country’s development of urbanization or income level reached a certain
threshold. The estimation results also confirmed that the effect of
renewable energy transition on environmental sustainability was
non-linear and varied significantly depending on a country’s level of
emissions. When per capita CO2 emissions of an economy are below the
first threshold, renewable energy transition leads to a decline in envi­
ronmental quality by increasing emissions. However, when per capita
CO2 emissions are above the first threshold, the effect of renewable
energy transition on environmental quality turns to positive, and it be­
comes even more profound when a country’s emission levels exceed the
second threshold. This finding can be explained by the fact that low per
capita CO2 emission countries in the LAC region are mainly low-income
countries, where per capita energy consumption is significantly lower
than in high-income countries due to a large share of population lacking
access to affordable and reliable energy services, who are unable to pay
for high energy bills because of their low purchasing power (Urban,
2014). As a result, a popular use of cheaper traditional biomass energy
sources (also known as conventional renewables), such as fuelwoods and
organic wastes for cooking and heating, is quite common in these
countries, ultimately leading to air pollution and environmental
degradation. On the contrary, countries with high per capita CO2
emissions in the LAC region are mostly upper-middle and high-income
nations characterized by considerably higher energy consumption
(both total and per capita) and larger production activities than their
low-income counterparts. Due to their superior macroeconomic and
fiscal capacities, high-income countries in the LAC region are better in
financing the high upfront costs required to carry out large-scale
deployment of renewable energy sources (utility-scale solar and wind
farms). As a result, the benefits from renewable energy transition in
high-income countries with high level of emissions and energy con­
sumption will be significantly greater than in low-income countries,
particularly in terms of reducing CO2 emissions. Dong et al. (2022. a), in
their study of 32 developed countries, also found a significant positive
relationship between renewable energy development and carbon emis­
sion efficiency when country’s income level was high.
Finally, the estimates of MMQR in this study confirm the heteroge­
nous impact of renewable energy transition and the digital economy on
economic growth and environmental sustainability in 18 LAC countries,
22
Journal of Cleaner Production 418 (2023) 138146
verifying that the validity of the Hypothesis 3 of this study also hold.
Unlike standard linear regressions and threshold regressions, the posi­
tive impact of renewable energy transition on economic growth is found
at all quantile groups. Its impact is stronger at lower quantiles and
gradually decreases across the quantile distribution. Renewable energy
transition is also found to have a positive and significant impact at the
1% significance level on environmental sustainability at all quantiles
and its impact gradually increases at higher quantiles. The results are
consistent with the study of Li et al. (2022b), who affirmed that
renewable energy had a beneficial impact on economic growth. Wang
et al. (2022a, 2022b, 2022c) also confirmed a positive effect of renew­
able energy on economic growth and its heterogenous impact in terms of
resource dependence and anticorruption regulation. With regard to the
digital economy, its impact on promoting economic growth is found at
the 1% significance level at all quantiles. Similar to renewable energy
transition, the digital economy has a strong impact at lower quantiles,
but its impact gradually decreases along the quantile distribution. It is
worth noting that the digital economy has a much stronger impact on
promoting economic growth than renewable energy transition in the
LAC region as the coefficients of LnDEI is systematically greater than
that of LnRETI across all quantile distribution. The results are consistent
with the findings of Hao et al. (2023), who found a significant positive
and spillover effects of digitalization on green economic growth in 277
cities in China. No evidence is found to support enough the robust
positive impact of the digital economy on improving environmental
quality in this study, as the coefficients of LnDEI are positive (more CO2
emissions) and statistically significant at the 10% significance level at
different quantile groups. Lange et al. (2020) reached a similar
conclusion that the digitalization might not contribute to reducing CO2
emissions due to a significant rebound effect and increased electricity
consumption by the ICT sectors. Main empirical findings of this study
are summarized with their impacts in Table 12. And the overview of this
research is illustrated in Fig. 2 by the categories: hwo to approach, what
are done, what outcomes are, and policy implications.
6. Conclusion and policy implications
In this study, the synergistic effect between renewable energy and
the digital economy was investigated in 18 Latin American and the
Caribbean countries during the period 2003–2019 from the perspective
of economic growth and environmental sustainability. New multidi­
mensional indices, the Renewable Energy Transition Index (RETI) and
the Digital Economy Index (DEI) specific to the LAC region were
developed to encompass multidimensional features and complex reality
accompanied by renewable energy transition and the digital economy in
evaluating the impact on green economic growth. Furthermore, the
FEPTR and the MMQR techniques were used to estimate the non-linear
and heterogenous impact of renewable energy transition and the digital
economy. Based on the estimation results, this study substantiated that
the synergistic effect improved environmental sustainability in various
quantile groups in the LAC region. However, the effect on boosting
economic growth could be observed only at the lowest-quantile (20th
quantile) using the MMQR. The robustness test was carried out by
incorporating time fixed effects into the MMQR and also confirmed the
validity of the estimation results. On the other hand, the FEPTR analysis
indicated that renewable energy transition had an important threshold
effect in association with the levels of income and CO2 emissions in the
LAC region. Regarding the threshold effect, renewable energy transition
did negatively affect economic growth when per capita GDP of the
country was above the first and second thresholds, and its negative
impact became larger as income level increased. Nevertheless, no sig­
nificant effect of renewable energy transition on economic growth was
found in low-income countries (i.e., those countries with per capita GDP
below the first threshold). For countries with per capita CO2 emissions
below the threshold, renewable energy transition led to their economic
growth, but meaningful impact was not confirmed when country’s
Y.K. Hwang
Table 12
Summary of empirical findings.
Countries Impacts confirmed
The LAC countries with 1) Strong synergistic effect between
relatively low levels RET and DE on promoting
of income economic growth at the 1% level
at Q20.
2) Strong effect of RET and DE on
promoting economic growth at
the 1% level at Q20 and Q40.
No significant effect of RET on
economic growth and CO2 emissions
below the first threshold.
The LAC countries with 1) Weak synergistic effect between
relatively low levels RET and the DE on improving
of CO2 emissions environmental sustainability at
the 10% level at Q40.a
2) Strong effect of RET on improving
environmental quality at the 1%
level at Q20 and Q40.
3) Weak negative effect of DE on
environmental quality at the 10%
level at Q20 and Q40.
1) Strong positive effect of RET on
economic growth at the 1% level
below the first threshold.
2) Strong negative effect of RET on
environmental quality at the 1%
level below the first threshold.
The LAC countries with 1) No synergistic effect between RET
relatively high levels and DE on promoting economic
of income growth.
2) Strong effect of RET and DE on
promoting economic growth at
the 1% level at Q60, Q80 and Q90
1) Strong negative effect of RET on
economic growth at the 1% level
above the first and the second
threshold (increasing effect)
2) Moderate positive effect of RET
on environmental quality at the
5% level above the first threshold.
The LAC countries with 1) Weak synergistic effect between
relatively high levels RET and the DE on improving
of CO2 emissions environmental sustainability at
the 10% level at Q60 and Q80.
2) Strong effect of RET on improving
environmental quality at the 1%
level at Q60,Q80, and Q90.
3) Negative effect of DE on
environmental quality at the 5%
and 10% level at Q60 and Q80
respectively.
1) No significant effect of RET on
economic growth above the first
threshold.
2) Positive effect of RET on
environmental quality
(increasing effect) at the 5% and
the 1% level above the first and
the second threshold respectively.
Note: DE: Digital Economy; FEPTR: Fixed Effects Panel Threshold Regression;
MMQR: Methods of Moments Quantile Regression; RET: Renewable Energy
Transition.
a
When the time fixed effects are not considered, the synergistic effect is found
at the 10% significance level at Q20 and Q40.
emissions levels were above the threshold. When it comes to the
threshold effect of renewable energy transition on environmental sus­
tainability, renewable energy transition had a positive effect only when
country’s income levels were above the threshold, but significant effect
was not found in the countries with income level below the threshold.
On the other hand, the negative effect of renewable energy transition
was found when the level of country’s CO2 emissions were below the
Journal of Cleaner Production 418 (2023) 138146
first threshold, but renewable energy transition led to improved envi­
ronmental quality when per capita CO2 emissions exceeded the first and
Reasoning basis second thresholds. This study also found the heterogenous impacts of
renewable energy transition and the digital economy in the LAC region.
MMQR with
time fixed Unlike the synergistic effect, the positive effect of renewable energy
effects transition on both economic growth and environmental sustainability
was verified across all quantile groups, and its impact was particularly
stronger at lower quantile. The digital economy had a positive effect on
economic growth in all quantile groups and its impact was much
FEPTR stronger at lower quantile. However, the digital economy did not lead to
improved environmental sustainability as it increased CO2 emissions in
some quantile groups.
MMQR with The following policy implications could be drawn from the empirical
time fixed results obtained in this research.
effects First, as far as the high potential benefits of the synergy between
renewable energy transition and the digital economy are concerned, the
governments in the LAC region should prioritize the integration of
renewable energy promoted by digitalization as urgent implementation
strategies in order to achieve a rapid decarbonization of their economies
and sustainable development in the region. The broad application of
FEPTR digital technologies in power grids dominated by renewables such as
smart grids can bring important energy saving costs thanks to more
efficient energy management and can provide more flexibility, reli­
ability, and resilience to energy systems alike by reducing a mismatch
between supply and demand and by accurate weather forecasting. Given
that many countries in the LAC region are suffering from a lack of effi­
MMQR with
cient energy system infrastructure, qualified workforce, and techno­
time fixed
effects logical and innovation capabilities to assimilate the advantages of
digitalization in power grids, consolidating regional energy system
integration and close collaboration between public and private sectors,
as well as establishing strong cooperation with the Global North (group
FEPTR
formed by advanced economies) can be a good alternative for the LAC
region to fully exploit the potential benefits from decarbonization and
digitalization process on economic growth because they can deliver
important energy security and cost savings thanks to more resilience to
price volatility of oil and gas, the economy of scale and can generate
spillover effects by promoting sharing of technical knowledge and skills
MMQR with as well as facilitating transfer of advanced green technology from
time fixed
effects
developed countries.
Second, given that each country in the LAC region has been facing
different challenges in terms of renewable energy transition and digi­
talization of their economy to achieve net-zero goals and sustainable
development goals, one-size-fits-all strategy is not adequate anymore.
Therefore, the governments in the region are encouraged to take feasible
measures adaptable to their specific context. Four different measures are
recommended, taking into consideration the country’s level of income
FEPTR and CO2 emissions: First, for the LAC countries with relatively lower
income levels, policymakers should exert great efforts to mobilize do­
mestic private capital to finance costly digitalization of renewable en­
ergy power system and to create favorable financial conditions to attract
large capital investments from foreign and domestic investors. Given the
fact that fiscal capacities of governments to finance high capital-
intensive renewable energy deployment and the development of digi­
tal infrastructures in these countries are quite limited, specific funds
devised by international development banks for clean energy deploy­
ment projects in developing countries such as green bonds can be of
great help. It is also recommended that the governments in low-income
countries in the LAC should proactively support and incentivize
renewable energy deployment and digitalization of their economies to
lower the high poverty level, accelerate economic growth, improve
living standards of their citizens, and reduce carbon emissions. Second,
for the LAC countries with relatively low levels of CO2 emissions, goal-
oriented policy measures should be taken by the governments to
strengthen their efforts towards a more sustainable pathways so that the
benefits of digitalization of energy mix can be fully exploited in terms of
environmental preservation. In this regard, it can be of great help to
23
Y.K. Hwang
Fig. 2. Summary of the reserch.
offer specific skill trainings and development programs to train highly
competitive workers for managing complex digital tools in advanced
renewable energy sector. The governments and policymakers in the LAC
countries with low level of CO2 emissions should pay careful attention to
avoid potential rebound effects due to digitalization. Third, for the LAC
countries with relatively high levels of income, governments should
implement policy measures to reduce their high dependence on hydro­
carbons in their economic structure and facilitate the integration of re­
newables in their energy mix in order to achieve Sustainable
Development Goals. To this end, governments can establish favorable
fiscal incentives such as tax credit to private investors and firms who
actively invest and adopt green energy technology in their production
line, offer Power Purchase Agreements (PPA) for low-carbon energy
producers to guarantee electricity prices in longer term and remove
political barriers which might hinder a prompt and successful renewable
energy transition such as vested interest of large oil and gas companies
in influencing political decisions in favor of them and streamline un­
necessary bureaucratic process to facilitate rapid renewable energy
deployment. Lastly, for the LAC countries with relatively high levels of
CO2 emissions, the policymakers should increase their efforts towards
decoupling of economic growth from environmental degradation with
special focus on energy-intensive and high-emitting sectors such as steel
and cement industries, transport sector which is heavily reliant on oil
and its derivatives as fuels for internal combustion engine cars. In this
respect, the use of green hydrogen and promotion of electromobility
(electric cars) can be a good alternative. The governments in these
countries also should take optimal measures to improve energy effi­
ciency and reduce energy consumption (product labelling to induce
sustainable production and consumption patterns, retrofit in buildings
and old power system infrastructure to increase their energy efficiency)
accompanied by an effective demand side management to induce con­
sumers’ behavioral change in order to optimize their energy use and
avoid potential rebound effects. It is also recommended that govern­
ments in the LAC countries with high level of CO2 emissions campaign to
raise public awareness of the seriousness of climate change and global
warming among citizens. These campaigns should be focused on
behavioral changes towards the use of cleaner energy sources for heat­
ing and cooking and restrain the public to use high-polluting biomass
24
Journal of Cleaner Production 418 (2023) 138146
energy sources.
Given that some of the important indicators of the digital economy in
the LAC region were unavailable during the research, such as medium
and high-tech manufacturing value added, Internet bandwidth and
speed, and diffusion rate of 5G networks, it was not possible to incor­
porate them into the design of the DEI. If these indicators were available,
we could make the DEI more concrete. It would also have been inter­
esting to investigate the effects of sub-dimensions of the RETI and DEI to
see what aspect of them has a greater influence on boosting economic
growth and improving environmental sustainability in the LAC region.
The study can be further extended to find the salient mechanisms
through which renewable energy transition and the digital economy can
be of a great contribution to green economic growth in the LAC region.
Furthermore, it will also be interesting to analyze the spatial spillover
effects of renewable energy transition and the digital economy on eco­
nomic growth and carbon emissions in the LAC countries in association
with their geographical location (country’s proximity to critical min­
erals reserves for clean energy transition and tec hubs).
Financial disclosure
None reported.
CRediT authorship contribution statement
Young Kyu Hwang: Conceptualization, Methodology, Software,
Writing – original draft, Writing – review & editing, Investigation,
Visualization.
Declaration of competing interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Data availability
Data will be made available on request.
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Appendix
Table A.0
Principal Factor Analysis

RETI

Factor Eigenvalue Difference Proportion Cumulative

Factor1 3.58625 1.50593 0.3260 0.3260

Factor2 2.08032 0.68130 0.1891 0.5151
Factor3 1.39903 0.21466 0.1272 0.6423
Factor4 1.18437 0.20917 0.1077 0.7500
DEI

Factor Eigenvalue Difference Proportion Cumulative

Factor1 1.84872 0.34834 0.1422 0.1422

Factor2 1.50038 0.04470 0.1154 0.2576
Factor3 1.45568 0.11883 0.1120 0.3696
Factor4 1.33685 0.21515 0.1028 0.4724
Factor5 1.12170 0.10554 0.0863 0.5587
Factor6 1.01617 0.03970 0.0782 0.6369

Table A.1
Correlation matrix

Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

(1) LnGDPpc 1.000

(2) LnCO2pc 0.736*** 1.000
(3) LnRETI 0.490*** 0.247*** 1.000
(4) LnDEI 0.329*** 0.203** 0.210** 1.000
(5) LnRETIxLnDEI − 0.020 − 0.101 − 0.025 − 0.010 1.000
(6) LnTNRR − 0.079 0.141 − 0.186* − 0.020 − 0.114 1.000
(7) LnInflation 0.153 0.149 − 0.045 0.004 0.021 0.030 1.000
(8) LnFDI 0.137 0.112 0.294*** 0.132 0.022 − 0.215*** − 0.006 1.000
(9) LnGI 0.668*** 0.537*** 0.524*** 0.229*** − 0.123 − 0.077 − 0.025 0.284*** 1.000
(10) LnAFFVadd − 0.736*** − 0.738*** − 0.628*** − 0.206** 0.039 0.120 0.039 − 0.191** − 0.582*** 1.000

Table A.2
VIF test

Model 1 Model 2

Variable VIF 1/VIF Variable VIF 1/VIF

LnRETI 1.47 0.681 LnRETI 1.47 0.681

LnGI 1.44 0.692 LnGI 1.47 0.681
LnFDI 1.16 0.859 LnFDI 1.16 0.859
LnDEI 1.07 0.933 LnTNRR 1.08 0.922
LnTNRR 1.07 0.934 LnDEI 1.07 0.933
LnInflation 1.00 0.997 LnRETIxLnDEI 1.03 0.968
Mean VIF 1.20 LnInflation 1.00 0.997
Mean VIF 1.19

Model 3 Model 4
Variable VIF 1/VIF Variable VIF 1/VIF

LnAFFVadd 1.94 0.514 LnAFFVadd 1.94 0.514

LnRETI 1.83 0.546 LnRETI 1.83 0.546
LnGI 1.68 0.595 LnGI 1.71 0.585
LnFDI 1.13 0.884 LnFDI 1.13 0.882
LnDEI 1.07 0.931 LnDEI 1.07 0.931
LnInflation 1.00 0.998 LnRETIxLnDEI 1.02 0.979
Mean VIF 1.44 LnInflation 1.00 0.997
Mean VIF 1.39
Note: Model 1–4 correspond to equation (1)- (4) respectively in section 4.4.

25
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table A.3
Cross-section dependence test

Variable CD-test p-value corr abs (corr)

LnGDPpc 38.44*** 0.000 0.754 0.869

LnCO2pc 12.07*** 0.000 0.237 0.509
LnTNRR 13.02*** 0.000 0.255 0.381
LnInflation 12.42*** 0.000 0.244 0.379
LnFDI 6.51*** 0.000 0.128 0.268
LnGI 32.76*** 0.000 0.642 0.676
LnRETI 6.87*** 0.000 0.135 0.497
LnDEI 6.00*** 0.000 0.118 0.313
LnRETIxLnDEI 1.45 0.147 0.028 0.310
LnAFFVadd 10.56*** 0.000 0.207 0.450
Note: ***, **, * denote statistical significance level at 1%,5%, and 10% respectively; H0 : cross-section independence.

Table A.4
Unit root test

Maddala and Wu Panel Unit Root test (MW)

At levels

Variable Specification without trend Specification with trend

Chi-sq p-value Chi-sq p-value

LnGDPpc 69.761*** 0.001 14.974 0.999

LnCO2pc 56.374** 0.017 25.202 0.911
LnTNRR 46.523 0.112 44.461 0.157
LnInflation 90.067*** 0.000 145.549*** 0.000
LnFDI 170.085*** 0.000 139.881*** 0.000
LnGI 184.769*** 0.000 54.990** 0.022
LnRETI 34.608 0.535 66.730*** 0.001
LnDEI 173.777*** 0.000 109.980*** 0.000
LnRETIxLnDEI 244.892*** 0.000 150.424*** 0.000
LnAFFVadd 64.335*** 0.003 82.204*** 0.000

At first difference
Variable Specification without trend Specification with trend
Chi-sq p-value Chi-sq p-value

DLnGDPpc 116.046*** 0.000 114.868*** 0.000

DLnCO2pc 193.850*** 0.000 158.079*** 0.000
DLnTNRR 212.432*** 0.000 171.350*** 0.000
DLnInflation 435.686*** 0.000 334.921*** 0.000
DLnFDI 482.563*** 0.000 400.023*** 0.000
DLnGI 184.801*** 0.000 247.834*** 0.000
DLnRETI 262.613*** 0.000 225.039*** 0.000
DLnDEI 264.095*** 0.000 183.361*** 0.000
DLnRETIxDLnDEI 274.792*** 0.000 196.996*** 0.000
DLnAFFVadd 284.125*** 0.000 226.529*** 0.000
Second generation Pesaran CIPS test

At levels

Variable Specification without trend Specification with trend

Zt-bar p-value Zt-bar p-value

LnGDPpc 1.884 0.970 4.718 1.000

LnCO2pc 0.328 0.628 2.318 0.990
LnTNRR − 0.679 0.249 0.385 0.650
LnInflation − 5.142*** 0.000 − 3.632*** 0.000
LnFDI − 5.155*** 0.000 − 4.210*** 0.000
LnGI − 1.307* 0.096 − 3.031*** 0.001
LnRETI − 1.144*** 0.126 − 1.757** 0.039
LnDEI − 6.236*** 0.000 − 4.197*** 0.000
LnRETIxLnDEI − 5.380*** 0.000 − 3.541*** 0.000
LnAFFVadd − 1.860** 0.031 − 0.169 0.433

At first difference
Variable Specification without trend Specification with trend
Zt-bar p-value Zt-bar p-value

DLnGDPpc − 3.120*** 0.001 − 2.602*** 0.005

DLnCO2pc − 7.246*** 0.000 − 6.601*** 0.000
DLnTNRR − 8.422*** 0.000 − 7.445*** 0.000
DLnInflation − 10.304*** 0.000 − 7.796*** 0.000
DLnFDI − 12.614*** 0.000 − 10.411*** 0.000
DLnGI − 10.933*** 0.000 − 8.846*** 0.000
DLnRETI − 9.038*** 0.000 − 7.516*** 0.000
(continued on next page)

26
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table A.4 (continued )

Second generation Pesaran CIPS test

At levels

Variable Specification without trend Specification with trend

Zt-bar p-value Zt-bar p-value

DLnDEI − 9.842*** 0.000 − 7.142*** 0.000

DLnRETIxDLnDEI − 9.192*** 0.000 − 6.459*** 0.000
DLnAFFVadd − 8.195*** 0.000 − 6.604*** 0.000
Note: ***, **, * denote statistical significance level at 1%,5%, and 10% respectively; H0 : series is I (1); the prefix D represents first dif­
ference.

Table A.5
Cointegration test

Pedroni cointegration test

Model 1 Model 2 Model 3 Model 4

Statistic Statistic Statistic Statistic

Modified Phillips–Perron t 6.8510*** 7.6340*** 6.6399*** 7.2500***

Phillips–Perron t 0.7732 0.4867 0.4252 − 0.8973
Augmented Dickey–Fuller t 1.6949** 1.2535 − 0.2237 − 0.9155

Westerlund cointegration test

Model 1 Model 2 Model 3 Model 4
Statistic Statistic Statistic Statistic

Variance ratio 7.1719*** 9.4704*** 2.4929*** 3.7122***

Note: ***, **, * denote statistical significance level at 1%,5%, and 10% respectively; Model 1–4 correspond to equations (1)–(4) respectively
in section 4.4; both trend and the option ‘all panels’ were included in Westerlund cointegration test. If the option ‘all panels’ was not set, it has
a Ha : cointegration of some panels, thus H0 :of no cointegration of all panels, Ha : cointegration of all panels.

Table A.6
Hausman test

Model 1 Model 2 Model 3 Model 4

Statistics chi2 (6) = 27.37*** chi2 (7) = 31.72*** chi2 (6) = 30.24*** chi2 (7) = 28.46***
Note: ***, **, * denote statistical significance level at 1%,5%, and 10% respectively; Model 1–4 correspond to equations (1)–(4) respectively in section 4.4; H0 :
not systematic difference in coefficients (presence of RE).

Table A.7
Breusch & Pagan Lagrangian multiplier test

Model 1 Model 2 Model 3 Model 4

Statistics chibar2 (01) = 1389.52*** chibar2 (01) = 1369.88*** chibar2 (01) = 830.90*** chibar2 (01) = 820.08***
Note: ***, **, * denote statistical significance level at 1%,5%, and 10% respectively; Model 1–4 correspond to equations (1)–(4) respectively in section 4.4; H0 : POLS is
preferred rather than static panel.

Table A.8
Wooldridge test

Model 1 Model 2 Model 3 Model 4

Statistics F (1,17) = 279.389*** F (1,17) = 296.260*** F (1,17) = 52.890*** F (1,17) = 55.322***

Note: ***, **, * denote statistical significance level at 1%,5%, and 10% respectively; Model 1–4 correspond to equations (1)–(4) respectively in section 4.4; H0 : no first-
order autocorrelation.

Table A.9
Modified Wald test for groupwise heteroskedasticity

Model 1 Model 2 Model 3 Model 4

Statistics chi2 (18) = 622.44*** chi2 (18) = 592.66*** chi2 (18) = 762.90*** chi2 (18) = 751.76***
Note: ***, **, * denote statistical significance level at 1%,5%, and 10% respectively; Model 1–4 correspond to equations (1)–(4) respectively in section 4.4; H0 : no
heteroskedasticity.

27
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table A.10
Temporal effect (Wald test)

Model 1 Model 2 Model 3 Model 4

Statistics F (16, 266) = 17.60*** F (16, 265) = 17.63*** F (16,266) = 1.72** F (16,265) = 1.71**
Note: ***, **, * denote statistical significance level at 1%,5%, and 10% respectively; Model 1–4 correspond to equation (1)- (4) respectively in section 4.4; H0 : time
fixed effect not significant.

FEPTR

• FEPTR with Yit = Ln GDPpcit , Rit = Ln RETIit , Qit = Ln GDPpcit .

Table A.11
Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln RETIit , Qit = Ln GDPpcit )

model Threshold Lower Upper

Th-1 9.362 9.305 9.370

Th-21 9.362 9.305 9.370
Th-22 8.530 8.521 8.534

Table A.12
Double Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln RETIit , Qit = Ln GDPpcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 1.157 0.004 57.200 0.133 62.431 73.869 89.254

Double 0.960 0.003 59.620 0.070 51.953 65.297 75.772

Table A.13
FEPTR estimation results (Yit = Ln GDPpcit , ; Rit = Ln RETIit , Qit = Ln GDPpcit )

LnGDPpc Coefficient Std.err t P>t [95% Confidence interval]

LnTNRR 0.039*** 0.010 3.880 0.000 0.019 0.059

LnInflation − 0.020** 0.010 − 2.050 0.041 − 0.038 − 0.001
LnFDI 0.009 0.007 1.280 0.203 − 0.005 0.023
LnGI 0.395*** 0.148 2.660 0.008 0.103 0.688
LnDEI − 0.028 0.031 − 0.900 0.367 − 0.089 0.033

_cat#c.LnRETI
0 0.000 0.029 0.010 0.991 − 0.056 0.057
1 − 0.086*** 0.031 − 2.730 0.007 − 0.148 − 0.024
2 − 0.245*** 0.036 − 6.760 0.000 − 0.316 − 0.173
_cons 6.780*** 0.608 11.150 0.000 5.582 7.977
sigma_u 0.542
sigma_e 0.060
rho 0.988 (Fraction of variance due to ui )
R2 (within) 0.837
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.

Table A.14
Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln RETIit , Qit = Ln GDPpcit )

model Threshold Lower Upper

Th-1 9.362 9.280 9.370

Th-21 9.362 9.305 9.370
Th-22 8.530 8.521 8.534

28
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table A.15
Double Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln RETIit , Qit = Ln GDPpcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 1.147 0.004 59.040 0.073 56.272 64.251 75.865

Double 0.960 0.003 56.470 0.057 46.759 58.537 75.224

Table A.16
FEPTR estimation results (Yit = Ln GDPpcit , ; Rit = Ln RETIit , Qit = Ln GDPpcit )

LnGDPpc Coefficient Std.err t P>t [95% Confidence interval]

LnTNRR 0.039 0.010 3.860 0.000 0.019 0.059

LnInflation − 0.020 0.010 − 2.050 0.042 − 0.038 − 0.001
LnFDI 0.009 0.007 1.270 0.204 − 0.005 0.023
LnGI 0.396 0.149 2.660 0.008 0.102 0.689
LnDEI − 0.028 0.031 − 0.900 0.368 − 0.089 0.033
LnRETIxLnDEI 0.002 0.091 0.020 0.983 − 0.177 0.181
_cat#c.LnRETI
0 0.000 0.029 0.010 0.990 − 0.056 0.057
1 − 0.086 0.032 − 2.720 0.007 − 0.148 − 0.024
2 − 0.245 0.036 − 6.750 0.000 − 0.316 − 0.173
_cons 6.779 0.611 11.100 0.000 5.577 7.981
sigma_u 0.542
sigma_e 0.061
rho 0.988 (Fraction of variance due to ui )
R2 (within) 0.837
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.
• FEPTR with Yit = Ln GDPpcit , Rit = Ln RETIit , Qit = Ln CO2pcit .

Table A.17
Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

model Threshold Lower Upper

Th-1 1.076 1.046 1.145

Th-21 1.076 1.046 1.145
Th-22 0.545 0.540 0.546

Table A.18
Double Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 1.093 0.004 77.490 0.010 45.986 57.257 75.783

Double 0.973 0.003 35.810 0.167 42.965 53.287 74.653

Table A.19
FEPTR estimation results (Yit = Ln GDPpcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

LnGDPpc Coefficient Std.err t P>t [95% Confidence interval]

LnTNRR 0.018 0.010 1.750 0.081 − 0.002 0.038

LnInflation − 0.006 0.009 − 0.640 0.524 − 0.025 0.013
LnFDI − 0.000 0.007 0.000 0.996 − 0.014 0.014
LnGI 0.280 0.146 1.910 0.057 − 0.008 0.569
LnDEI 0.005 0.031 0.160 0.873 − 0.056 0.066

_cat#c.LnRETI
0 0.132 0.036 3.710 0.000 0.062 0.202
1 0.038 0.029 1.300 0.194 − 0.019 0.095
2 − 0.245 0.043 − 5.680 0.000 − 0.330 − 0.160
_cons 7.343 0.602 12.190 0.000 6.157 8.529
sigma_u 0.515
sigma_e 0.061
rho 0.986 (Fraction of variance due to ui )
R2 (within) 0.836
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.

29
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table A.20
Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

model Threshold Lower Upper

Th-1 1.076 1.046 1.145

Th-21 1.076 1.046 1.145
Th-22 0.545 0.540 0.546

Table A.21
Double Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 1.091 0.004 76.930 0.010 49.926 58.093 72.051

Double 0.969 0.003 36.490 0.147 42.008 50.710 72.484

Table A.22
FEPTR estimation results (Yit = Ln GDPpcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

LnGDPpc Coefficient Std.err t P>t [95% Confidence interval]

LnTNRR 0.017 0.010 1.720 0.086 − 0.002 0.037

LnInflation − 0.005 0.010 − 0.540 0.591 − 0.024 0.014
LnFDI 0.000 0.007 0.010 0.994 − 0.014 0.014
LnGI 0.267 0.147 1.820 0.070 − 0.022 0.557
LnDEI 0.005 0.031 0.160 0.871 − 0.056 0.066
LnRETIxLnDEI − 0.093 0.089 − 1.050 0.294 − 0.268 0.081
_cat#c.LnRETI
0 0.130 0.036 3.650 0.000 0.060 0.200
1 0.035 0.029 1.210 0.227 − 0.022 0.093
2 − 0.246 0.043 − 5.700 0.000 − 0.331 − 0.161
_cons 7.395 0.604 12.240 0.000 6.206 8.585
sigma_u 0.516
sigma_e 0.061
rho 0.986 (Fraction of variance due to ui )
R2 (within) 0.836
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.
• FEPTR with Yit = Ln CO2pcit , Rit = Ln RETIit , Qit = Ln GDPpcit .

Table A.23
Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln GDPpcit )

model Threshold Lower Upper

Th-1 8.516 8.511 8.518

Th-21 8.516 8.511 8.518
Th-22 7.824 7.810 7.830

Table A.24
Double Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln GDPpcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 2.843 0.010 70.110 0.007 47.235 54.830 66.128

Double 2.615 0.009 25.190 0.410 40.608 47.875 68.522

Table A.25
FEPTR estimation results (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln GDPpcit )

LnCO2pc Coefficient Std.err t P>t [95% Confidence interval]

LnAFFVadd − 0.240 0.047 − 5.100 0.000 − 0.332 − 0.147

LnInflation − 0.005 0.016 − 0.290 0.774 − 0.035 0.026
LnFDI 0.031 0.012 2.640 0.009 0.008 0.055
LnGI − 0.065 0.236 − 0.270 0.784 − 0.530 0.401
LnDEI − 0.008 0.052 − 0.160 0.874 − 0.111 0.094
(continued on next page)

30
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table A.25 (continued )

LnCO2pc Coefficient Std.err t P>t [95% Confidence interval]

_cat#c.LnRETI
0 0.046 0.069 0.670 0.500 − 0.089 0.181
1 − 0.102 0.049 − 2.070 0.039 − 0.199 − 0.005
2 − 0.275 0.054 − 5.060 0.000 − 0.382 − 0.168
_cons 1.014 0.992 1.020 0.307 − 0.938 2.966
sigma_u 0.391
sigma_e 0.101
rho 0.938 (Fraction of variance due to ui )
R2 (within) 0.503
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.

Table A.26
Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln GDPpcit )

model Threshold Lower Upper

Th-1 8.516 8.511 8.518

Th-21 8.516 8.511 8.518
Th-22 7.824 7.810 7.830

Table A.27
Double Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln GDPpcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 2.837 0.010 69.610 0.020 44.686 55.718 71.540

Double 2.614 0.009 24.670 0.433 40.632 48.006 62.463

Table A.28
FEPTR estimation results (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln GDPpcit )

LnCO2pc Coefficient Std.err t P>t [95% Confidence interval]

LnAFFVadd − 0.238 0.047 − 5.050 0.000 − 0.331 − 0.145

LnInflation − 0.005 0.016 − 0.310 0.754 − 0.036 0.026
LnFDI 0.031 0.012 2.640 0.009 0.008 0.055
LnGI − 0.055 0.238 − 0.230 0.817 − 0.523 0.413
LnDEI − 0.009 0.052 − 0.180 0.860 − 0.112 0.094
LnRETIxLnDEI 0.069 0.152 0.460 0.647 − 0.229 0.368
_cat#c.LnRETI
0 0.046 0.069 0.670 0.502 − 0.089 0.182
1 − 0.101 0.049 − 2.040 0.043 − 0.198 − 0.003
2 − 0.276 0.054 − 5.060 0.000 − 0.383 − 0.169
_cons 0.969 0.998 0.970 0.332 − 0.996 2.934
sigma_u 0.391
sigma_e 0.101
rho 0.937 (Fraction of variance due to ui )
R2 (within) 0.503
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.
• FEPTR with Yit = Ln CO2pcit , Rit = Ln RETIit , Qit = Ln CO2pcit .

Table A.29
Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

model Threshold Lower Upper

Th-1 0.302 0.298 0.326

Th-21 0.123 0.121 0.151
Th-22 1.153 1.095 1.210

31
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table A.30
Double Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 2.737 0.009 84.010 0.020 47.540 61.671 92.674

Double 2.103 0.007 87.150 0.000 26.227 29.826 36.125

Table A.31
FEPTR estimation results (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

LnCO2pc Coefficient Std.err t P>t [95% Confidence interval]

LnAFFVadd − 0.144 0.043 − 3.370 0.001 − 0.228 − 0.060

LnInflation − 0.018 0.014 − 1.270 0.205 − 0.045 0.010
LnFDI 0.000 0.011 0.040 0.971 − 0.021 0.022
LnGI − 0.050 0.206 − 0.240 0.810 − 0.455 0.355
LnDEI 0.001 0.045 0.030 0.979 − 0.088 0.090

_cat#c.LnRETI
0 0.142 0.050 2.850 0.005 0.044 0.240
1 − 0.103 0.042 − 2.430 0.016 − 0.186 − 0.020
2 − 0.575 0.064 − 8.960 0.000 − 0.702 − 0.449
_cons 0.906 0.868 1.040 0.298 − 0.804 2.616
sigma_u 0.282
sigma_e 0.089
rho 0.909 (Fraction of variance due to ui )
R2 (within) 0.610
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.

Table A.32
Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

model Threshold Lower Upper

Th-1 0.302 0.298 0.326

Th-21 0.123 0.121 0.151
Th-22 1.153 1.095 1.210

Table A.33
Double Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 2.707 0.009 86.800 0.007 42.760 53.294 78.122

Double 2.093 0.007 84.810 0.000 25.643 31.131 41.783

Table A.34
FEPTR estimation results (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

LnCO2pc Coefficient Std.err t P>t [95% Confidence interval]

LnAFFVadd − 0.147 0.043 − 3.450 0.001 − 0.232 − 0.063

LnInflation − 0.016 0.014 − 1.170 0.243 − 0.043 0.011
LnFDI 0.000 0.011 0.040 0.965 − 0.021 0.022
LnGI − 0.076 0.207 − 0.370 0.713 − 0.484 0.331
LnDEI 0.001 0.045 0.030 0.979 − 0.088 0.090
LnRETIxLnDEI − 0.147 0.130 − 1.130 0.260 − 0.403 0.109
_cat#c.LnRETI
0 0.138 0.050 2.760 0.006 0.040 0.237
1 − 0.107 0.042 − 2.530 0.012 − 0.191 − 0.024
2 − 0.578 0.064 − 9.010 0.000 − 0.705 − 0.452
_cons 1.019 0.874 1.170 0.244 − 0.701 2.739
sigma_u 0.283
sigma_e 0.089
rho 0.909 (Fraction of variance due to ui )
R2 (within) 0.612
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.

32
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

• FEPTR with Yit = Ln GDPpcit , Rit = Ln REShareTFECit , Qit = Ln GDPpcit .

Table A.35
Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln REShareTFECit , Qit = Ln GDPpcit )

model Threshold Lower Upper

Th-1 9.481 9.460 9.481

Th-21 9.454 9.450 9.466
Th-22 8.580 8.577 8.581

Table A.36
Double Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln REShareTFECit , Qit = Ln GDPpcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 0.865 0.003 107.730 0.003 53.284 65.879 92.792

Double 0.686 0.002 75.330 0.007 48.880 56.280 69.400

Table A.37
FEPTR estimation results (Yit = Ln GDPpcit , ; Rit = Ln REShareTFECit , Qit = Ln GDPpcit )

LnGDPpc Coefficient Std.err t P>t [95% Confidence interval]

LnTNRR 0.038 0.009 4.480 0.000 0.021 0.055

LnInflation − 0.006 0.008 − 0.680 0.500 − 0.022 0.011
LnFDI − 0.001 0.006 − 0.150 0.881 − 0.013 0.011
LnGI 0.428 0.125 3.430 0.001 0.182 0.674
LnDEI 0.016 0.026 0.620 0.537 − 0.035 0.068

_cat#c.LnREShareTFEC
0 − 0.207 0.021 − 10.050 0.000 − 0.247 − 0.166
1 − 0.173 0.021 − 8.240 0.000 − 0.215 − 0.132
2 − 0.118 0.021 − 5.580 0.000 − 0.159 − 0.076
Constant 7.346 0.500 14.700 0.000 6.362 8.330
Sigma u 0.509
Sigma e 0.051
rho 0.990 (Fraction of variance due to ui )
R2 (within) 0.883
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.

Table A.38
Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln REShareTFECit , Qit = Ln GDPpcit )

model Threshold Lower Upper

Th-1 9.481 9.460 9.481

Th-21 9.454 9.450 9.466
Th-22 8.580 8.577 8.581

Table A.39
Double Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln REShareTFECit , Qit = Ln GDPpcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 0.856 0.003 111.130 0.010 60.167 73.671 108.201

Double 0.682 0.002 73.590 0.007 45.576 54.512 72.747

Table A40
FEPTR estimation results (Yit = Ln GDPpcit , ; Rit = Ln REShareTFECit , Qit = Ln GDPpcit )

LnGDPpc Coefficient Std.err t P>t [95% Confidence interval]

LnTNRR 0.038 0.009 4.410 0.000 0.021 0.054

LnInflation − 0.005 0.008 − 0.670 0.505 − 0.021 0.011
LnFDI − 0.000 0.006 − 0.070 0.941 − 0.012 0.011
(continued on next page)

33
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table A40 (continued )

LnGDPpc Coefficient Std.err t P>t [95% Confidence interval]

LnGI 0.446 0.125 3.550 0.000 0.199 0.693

LnDEI 0.028 0.028 1.010 0.313 − 0.026 0.082
LnREShareTFECxLnDEI − 0.053 0.040 − 1.340 0.181 − 0.131 0.025
_cat#c.LnREShareTFEC
0 − 0.207 0.021 − 10.090 0.000 − 0.248 − 0.167
1 − 0.174 0.021 − 8.290 0.000 − 0.216 − 0.133
2 − 0.118 0.021 − 5.610 0.000 − 0.159 − 0.077
Constant 7.283 0.501 14.530 0.000 6.296 8.270
Sigma u 0.508
Sigma e 0.051
rho 0.990 (Fraction of variance due to ui )
R2 (within) 0.884
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.
• FEPTR with Yit = Ln GDPpcit , Rit = Ln REShareTFECit , Qit = Ln CO2pcit .

Table A.41
Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln REShareTFECit , Qit = Ln CO2pcit )

model Threshold Lower Upper

Th-1 0.545 0.542 0.546

Th-21 0.545 0.542 0.546
Th-22 1.346 1.323 1.349

Table A.42
Double Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln REShareTFECit , Qit = Ln CO2pcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 1.079 0.004 29.190 0.473 50.814 60.320 73.644

Double 0.948 0.003 39.740 0.120 41.685 49.629 67.403

Table A.43
FEPTR estimation results (Yit = Ln GDPpcit , ; Rit = Ln REShareTFECit , Qit = Ln CO2pcit )

LnGDPpc Coefficient Std.err t P>t [95% Confidence interval]

LnTNRR 0.024 0.010 2.430 0.016 0.005 0.043

LnInflation − 0.005 0.010 − 0.560 0.576 − 0.024 0.013
LnFDI − 0.005 0.007 − 0.680 0.497 − 0.019 0.009
LnGI 0.640 0.147 4.340 0.000 0.349 0.930
LnDEI 0.019 0.031 0.610 0.540 − 0.042 0.080

_cat#c.LnREShareTFEC
0 − 0.136 0.025 − 5.500 0.000 − 0.184 − 0.087
1 − 0.103 0.026 − 4.000 0.000 − 0.154 − 0.052
2 − 0.058 0.029 − 2.050 0.042 − 0.115 − 0.002
_cons 6.263 0.601 10.420 0.000 5.079 7.447
sigma_u 0.528
sigma_e 0.060
rho 0.987 (Fraction of variance due to ui )
R2 (within) 0.834
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.

Table A.44
Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln REShareTFECit , Qit = Ln CO2pcit )

model Threshold Lower Upper

Th-1 0.545 0.542 0.546

Th-21 0.545 0.542 0.546
Th-22 1.346 1.323 1.349

34
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table A.45
Double Threshold effect test (Yit = Ln GDPpcit , ; Rit = Ln REShareTFECit , Qit = Ln CO2pcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 1.077 0.004 29.090 0.483 55.697 67.590 86.996

Double 0.946 0.003 40.000 0.090 37.522 49.941 78.450

Table A.46
FEPTR estimation results (Yit = Ln GDPpcit , ; Rit = Ln REShareTFECit , Qit = Ln CO2pcit )

LnGDPpc Coefficient Std.err t P>t [95% Confidence interval]

LnTNRR 0.024 0.010 2.380 0.018 0.004 0.043

LnInflation − 0.005 0.010 − 0.550 0.586 − 0.024 0.014
LnFDI − 0.004 0.007 − 0.630 0.532 − 0.018 0.009
LnGI 0.653 0.148 4.410 0.000 0.362 0.945
LnDEI 0.029 0.033 0.880 0.380 − 0.035 0.093
LnREShareTFECxLnDEI − 0.045 0.047 − 0.950 0.341 − 0.136 0.047
_cat#c.LnREShareTFEC
0 − 0.136 0.025 − 5.510 0.000 − 0.184 − 0.087
1 − 0.103 0.026 − 4.010 0.000 − 0.154 − 0.053
2 − 0.059 0.029 − 2.050 0.042 − 0.115 − 0.002
_cons 6.214 0.604 10.290 0.000 5.025 7.402
sigma_u 0.527
sigma_e 0.060
rho 0.987 (Fraction of variance due to ui )
R2 (within) 0.838
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.
• FEPTR with Yit = Ln CO2pcit , Rit = Ln REShareTFECit , Qit = Ln GDPpcit .

Table A.47
Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln REShareTFECit , Qit = Ln GDPpcit )

model Threshold Lower Upper

Th-1 8.516 8.514 8.518

Th-21 8.530 8.519 8.541
Th-22 9.393 9.357 9.395

Table A.48
Double Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln REShareTFECit , Qit = Ln GDPpcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 1.800 0.006 68.940 0.020 47.942 60.497 73.508

Double 1.611 0.006 33.820 0.260 45.941 54.995 76.751

Table A.49
FEPTR estimation results (Yit = Ln CO2pcit , ; Rit = Ln REShareTFECit , Qit = Ln GDPpcit )

LnCO2pc Coefficient Std.err t P>t [95% Confidence interval]

LnAFFVadd − 0.095 0.039 − 2.400 0.017 − 0.172 − 0.017

LnInflation 0.011 0.012 0.850 0.398 − 0.014 0.035
LnFDI 0.021 0.009 2.310 0.022 0.003 0.039
LnGI − 0.229 0.181 − 1.270 0.207 − 0.585 0.127
LnDEI − 0.008 0.040 − 0.200 0.843 − 0.087 0.071

_cat#c.LnREShareTFEC
0 − 0.444 0.032 − 14.000 0.000 − 0.506 − 0.381
1 − 0.390 0.033 − 11.820 0.000 − 0.456 − 0.325
2 − 0.351 0.033 − 10.510 0.000 − 0.416 − 0.285
_cons 2.920 0.748 3.900 0.000 1.446 4.394
sigma_u 0.268
sigma_e 0.079
rho 0.921 (Fraction of variance due to ui )
R2 (within) 0.698
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.

35
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table A.50
Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln REShareTFECit , Qit = Ln GDPpcit )

model Threshold Lower Upper

Th-1 8.516 8.514 8.518

Th-21 8.516 8.514 8.518
Th-22 9.393 9.357 9.395

Table A.51
Double Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln REShareTFECit , Qit = Ln GDPpcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 1.795 0.006 69.810 0.017 47.893 57.812 78.015

Double 1.608 0.006 33.570 0.177 40.022 46.213 56.286

Table A.52
FEPTR estimation results (Yit = Ln CO2pcit , ; Rit = Ln REShareTFECit , Qit = Ln GDPpcit )

LnCO2pc Coefficient Std.err t P>t [95% Confidence interval]

LnAFFVadd − 0.100 0.039 − 2.550 0.011 − 0.178 − 0.023

LnInflation 0.011 0.012 0.860 0.392 − 0.014 0.035
LnFDI 0.022 0.009 2.370 0.019 0.004 0.040
LnGI − 0.239 0.181 − 1.320 0.188 − 0.594 0.117
LnDEI − 0.021 0.042 − 0.490 0.623 − 0.103 0.062
LnREShareTFECxLnDEI 0.050 0.061 0.820 0.411 − 0.069 0.169
_cat#c.LnREShareTFEC
0 − 0.444 0.032 − 14.070 0.000 − 0.506 − 0.382
1 − 0.389 0.033 − 11.820 0.000 − 0.454 − 0.324
2 − 0.350 0.033 − 10.560 0.000 − 0.416 − 0.285
_cons 2.960 0.746 3.970 0.000 1.490 4.430
sigma_u 0.265
sigma_e 0.078
rho 0.920 (Fraction of variance due to ui )
R2 (within) 0.702
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.
• FEPTR with Yit = Ln CO2pcit , Rit = Ln REShareTFECit , Qit = Ln CO2pcit .

Table A.53
Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

model Threshold Lower Upper

Th-1 − 0.001 − 0.009 0.019

Th-21 − 0.001 − 0.009 0.019
Th-22 0.518 0.505 0.522

Table A.54
Double Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln REShareTFECit , Qit = Ln CO2pcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 1.810 0.006 66.990 0.013 44.147 51.234 67.055

Double 1.542 0.005 50.100 0.037 41.414 48.368 67.598

Table A.55
FEPTR estimation results (Yit = Ln CO2pcit , ; Rit = Ln REShareTFECit , Qit = Ln CO2pcit )

LnCO2pc Coefficient Std.err t P>t [95% Confidence interval]

LnAFFVadd − 0.198 0.036 − 5.530 0.000 − 0.268 − 0.127

LnInflation 0.011 0.012 0.890 0.372 − 0.013 0.035
LnFDI 0.014 0.009 1.520 0.129 − 0.004 0.031
LnGI − 0.070 0.171 − 0.410 0.683 − 0.406 0.267
LnDEI 0.011 0.039 0.270 0.785 − 0.066 0.088
(continued on next page)

36
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table A.55 (continued )

LnCO2pc Coefficient Std.err t P>t [95% Confidence interval]

cat#c.LnREShareTFEC
0 − 0.516 0.031 − 16.580 0.000 − 0.577 − 0.455
1 − 0.465 0.030 − 15.270 0.000 − 0.525 − 0.405
2 − 0.421 0.031 − 13.430 0.000 − 0.483 − 0.360
_cons 2.659 0.710 3.740 0.000 1.261 4.058
sigma_u 0.243
sigma_e 0.077
rho 0.910 (Fraction of variance due to ui )
R2 (within) 0.713
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.

Table A.56
Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln RETIit , Qit = Ln CO2pcit )

model Threshold Lower Upper

Th-1 − 0.001 − 0.009 0.019

Th-21 − 0.083 − 0.104 − 0.068
Th-22 0.518 0.505 0.522

Table A.57
Double Threshold effect test (Yit = Ln CO2pcit , ; Rit = Ln REShareTFECit , Qit = Ln CO2pcit )

Threshold RSS MSE Fstat Prob Crit10 Crit5 Crit1

Single 1.804 0.006 67.930 0.020 44.276 51.801 76.854

Double 1.532 0.005 51.360 0.037 41.025 48.744 62.184

Table A.58
FEPTR estimation results (Yit = Ln CO2pcit , ; Rit = Ln REShareTFECit , Qit = Ln CO2pcit )

LnCO2pc Coefficient Std.err t P>t [95% Confidence interval]

LnAFFVadd − 0.170 0.036 − 4.720 0.000 − 0.241 − 0.099

LnInflation 0.013 0.012 1.100 0.274 − 0.011 0.037
LnFDI 0.015 0.009 1.710 0.088 − 0.002 0.033
LnGI − 0.048 0.171 − 0.280 0.779 − 0.385 0.289
LnDEI − 0.006 0.041 − 0.140 0.888 − 0.087 0.075
LnREShareTFECxLnDEI − 0.084 0.060 − 1.390 0.166 − 0.203 0.035
cat#c.LnREShareTFEC
0 − 0.532 0.032 − 16.870 0.000 − 0.594 − 0.470
1 − 0.465 0.030 − 15.270 0.000 − 0.524 − 0.405
2 − 0.420 0.031 − 13.400 0.000 − 0.482 − 0.358
_cons 2.474 0.710 3.480 0.001 1.076 3.872
sigma_u 0.248
sigma_e 0.076
rho 0.913 (Fraction of variance due to ui )
R2 (within) 0.715
Number of obs. 306
Note: *, **, and *** denote significance level at 10%, 5%, and 1% respectively and time dummies were included during the estimation.

MMQR
Table A.59
Test for slope homogeneity for the MMQR

Statistics MMQR 1 MMQR 2

Delta (Hac) Delta (Hac)adj Delta (Hac) Delta (Hac)adj

− 3.522*** − 4.840*** − 4.176*** − 6.088***

Statistics MMQR 3 MMQR 4

Delta (Hac) Delta (Hac)adj Delta (Hac) Delta (Hac)adj

− 2.924*** − 4.018*** − 3.704*** − 5.400***

Note: ***, **, * denote statistical significance level at 1%,5%, and 10% respectively; MMQR 1- MMQR 4 correspond to equation 12–15
respectively in section 4.6; the option ‘Hac’ which refers to heteroskedasticity and autocorrelation consistent was included in the test;
H0 : slope coefficients are homogenous.

37
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Table A.60
Shapiro–Wilk W test for normal data for the MMQR

Variable Obs W V z Prob > z

LnGDPpc 306 0.9658 7.405 4.704*** 0.000

LnCO2pc 306 0.9631 8.000 4.886*** 0.000
Note: ***, **, * denote statistical significance level at 1%,5%, and 10% respectively; H0 : normality of the data.

Table A.61
Skewness-Kurtosis test for the MMQR

Variable Obs Pr (skewness) Pr (kurtosis) Joint test

Adj chi2 (2) Prob > chi2

LnGDPpc 306 0.0735 0.0000 26.04 0.000

LnCO2pc 306 0.5198 0.0000 39.04 0.000
Note: ***, **, * denote statistical significance level at 1%,5%, and 10% respectively; H0 : normality of the data.

Fig. A1. Kernel and normal density for LnGDPpc and LnCO2pc

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jclepro.2023.138146.

References organisation for economic Co-operation and development countries. J. Clean. Prod.
242, 118537 https://doi.org/10.1016/j.jclepro.2019.118537.
Dong, F., Li, Y., Gao, Y., Zhu, J., Qin, C., Zhang, X., 2022 a. Energy transition and carbon
Alola, A.A., Bekun, F.V., Sarkodie, S.A., 2019. Dynamic impact of trade policy, economic
neutrality: exploring the non-linear impact of renewable energy development on
growth, fertility rate, renewable and non-renewable energy consumption on
carbon emission efficiency in developed countries. Resour. Conserv. Recycl. 177,
ecological footprint in Europe. Sci. Total Environ. 685, 702–709.
106002 https://doi.org/10.1016/j.resconrec.2021.106002.
Alvarado, R., Tillaguango, B., Dagar, V., Ahmad, M., Işık, C., Méndez, P., Toledo, E.,
Dong, F., Li, Y., Qin, C., Zhang, X., Chen, Y., Zhao, X., Wang, C., 2022 b. Information
2021. Ecological footprint, economic complexity, and natural resources rents in
infrastructure and greenhouse gas emission performance in urban China: a
Latin America: empirical evidence using quantile regressions. J. Clean. Prod. 318,
difference-in-differences analysis. J. Environ. Manag. 316, 115252 https://doi.org/
128585.
10.1016/j.jenvman.2022.115252.
Apergis, N., Payne, J.E., 2010. Renewable energy consumption and economic growth:
Dong, K., Hochman, G., Zhang, Y., Sun, R., Li, H., Liao, H., 2018. CO2 emissions,
evidence from a panel of OECD countries. Energy Pol. 38 (1), 656–660. https://doi.
economic and population growth, and renewable energy: empirical evidence across
org/10.1016/j.enpol.2009.09.002.
regions. Energy Econ. 75, 180–192. https://doi.org/10.1016/j.eneco.2018.08.017.
Asif, M., 2020. Role of energy conservation and management in the 4D sustainable
Doytch, N., Narayan, S., 2016. Does FDI influence renewable energy consumption? An
energy transition. Sustainability 12 (23), 10006.
analysis of sectoral FDI impact on renewable and non-renewable industrial energy
Bárcena Ibarra, A., Samaniego, J., Peres, W., Alatorre, J.E., 2020. The Climate
consumption. Energy Econ. 54, 291–301.
Emergency in Latin America and the Caribbean: the Path Ahead–Resignation or
Doytch, N., Narayan, S., 2021. Does transitioning towards renewable energy accelerate
Action. ECLAC.
economic growth? An analysis of sectoral growth for a dynamic panel of countries.
Ben Jebli, M., Ben Youssef, S., Apergis, N., 2019. The dynamic linkage between
Energy 235, 121290. https://doi.org/10.1016/j.energy.2021.121290.
renewable energy, tourism, CO2 emissions, economic growth, foreign direct
ECLAC, 2022. A Digital Path for Sustainable Development in Latin America and the
investment, and trade. Latin American Economic Review 28 (1), 1–19.
Caribbean.
Bhattacharya, M., Paramati, S.R., Ozturk, I., Bhattacharya, S., 2016. The effect of
Fankhauser, S., Jotzo, F., 2018. Economic growth and development with low-carbon
renewable energy consumption on economic growth: evidence from top 38
energy. Wiley Interdisciplinary Reviews. Climate Change 9 (1). https://doi.org/
countries. Appl. Energy 162, 733–741. https://doi.org/10.1016/j.
10.1002/wcc.495 n/a.
apenergy.2015.10.104.
Fattouh, B., Poudineh, R., West, R.R., 2019. The Rise of Renewables and Energy
CEPAL, N., 2022 a. Latin American Economic Outlook 2022: towards a Green and Just
Transition: what Adaptation Strategy Exists for Oil Companies and Oil-Exporting
Transition. ECLAC.
Countries. Springer Science and Business Media LLC. https://doi.org/10.1007/
CEPAL, N., 2022 b. A Digital Path for Sustainable Development in Latin America and the
s41825-019-00013-x.
Caribbean.
Fuinhas, J.A., Marques, A.C. (Eds.), 2019. The Extended Energy–Growth Nexus: Theory
Charfeddine, L., Kahia, M., 2019. Impact of renewable energy consumption and financial
and Empirical Applications. Academic Press.
development on CO2 emissions and economic growth in the MENA region: a panel
Fuinhas, J.A., Koengkan, M., Santiago, R., 2021. Physical Capital Development and
vector autoregressive (PVAR) analysis. Renew. Energy 139, 198–213. https://doi.
Energy Transition in Latin America and the Caribbean. Elsevier.
org/10.1016/j.renene.2019.01.010.
Grossman, G.M., Krueger, A.B., 1995. Economic growth and the environment. Q. J. Econ.
D’agostino, R.B., Belanger, A., D’Agostino Jr., R.B., 1990. A suggestion for using
110 (2), 353–377.
powerful and informative tests of normality. Am. Statistician 44 (4), 316–321.
Grottera, C., 2022. Reducing Emissions from the Energy Sector for a More Resilient and
Destek, M.A., Sinha, A., 2020. Renewable, non-renewable energy consumption,
Low-Carbon Post-pandemic Recovery in Latin America and the Caribbean.
economic growth, trade openness and ecological footprint: evidence from

38
Y.K. Hwang
Guimarães, L.N. (Ed.), 2020. The Regulation and Policy of Latin American Energy
Transitions. Elsevier.
Hafner, M., Luciani, G., 2022. The Palgrave Handbook of International Energy
Economics. Springer Nature, p. 770.
Hampl, N., 2022. Equitable Energy Transition in Latin America and the Caribbean:
Reducing Inequity by Building Capacity. Elsevier BV. https://doi.org/10.1016/j.
rset.2022.100035.
Hao, X., Li, Y., Ren, S., Wu, H., Hao, Y., 2023. The role of digitalization on green
economic growth: does industrial structure optimization and green innovation
matter? J. Environ. Manag. 325, 116504 https://doi.org/10.1016/j.
jenvman.2022.116504.
Henderson, J., Sen, A., 2021. The Energy Transition: Key Challenges for Incumbent and New
Players in the Global Energy System (No. 01). OIES Paper: ET.
Hübner, K. (Ed.), 2018. National Pathways to Low Carbon Emission Economies:
Innovation Policies for Decarbonizing and Unlocking. Routledge.
IEA, 2017. Digitalization & Energy.
Inglesi-Lotz, R., 2016. The impact of renewable energy consumption to economic
growth: a panel data application. Energy Econ. 53, 58–63. https://doi.org/10.1016/
j.eneco.2015.01.003.
IPCC, 2022. Climate Change 2022: Impacts, Adaptation, and Vulnerability. Contribution of
Working Group II to the Sixth Assessment Report of the Intergovernmental Panel on
Climate Change. Cambridge University Press. Cambridge University Press,
Cambridge, UK and New York, NY, USA, p. 3056. https://doi.org/10.1017/
9781009325844.
Ivanovski, K., Hailemariam, A., Smyth, R., 2021. The effect of renewable and non-
renewable energy consumption on economic growth: non-parametric evidence.
J. Clean. Prod. 286, 124956 https://doi.org/10.1016/j.jclepro.2020.124956.
Jimenez, V.M.M., Gonzalez, E.P., 2022. The role of artificial intelligence in Latin
americas energy transition. IEEE Latin America Transactions 20 (11), 2404–2412.
Kaya, Y., 1990. Impact of Carbon Dioxide Emission Control on GNP Growth:
Interpretation of Proposed Scenarios. Paper presented to the IPCC Energy and In-
dustry Subgroup, Response Strategies Working Group, Paris. Mimeo.
Koengkan, M., Fuinhas, J.A., 2022. Globalisation and Energy Transition in Latin America
and the Caribbean. Springer Books.
Kuc-Czarnecka, M.E., Olczyk, M., Zinecker, M., 2021. Improvements and spatial
dependencies in energy transition measures. Energies 14 (13), 3802. https://doi.
org/10.3390/en14133802.
Labra, R., Torrecillas, C., 2014. Guía CERO para datos de panel. Un enfoque práctico.
UAM-Accenture Working Papers 16 (1), 57.
Lange, S., Pohl, J., Santarius, T., 2020. Digitalization and energy consumption. Does ICT
reduce energy demand? Ecol. Econ. 176, 106760 https://doi.org/10.1016/j.
ecolecon.2020.106760.
Li, J., Chen, L., Chen, Y., He, J., 2022 a. Digital economy, technological innovation, and
green economic efficiency—empirical evidence from 277 cities in China. Manag.
Decis. Econ. 43 (3), 616–629. https://doi.org/10.1002/mde.3406.
Li, R., Wang, X., Wang, Q., 2022 b. Does renewable energy reduce ecological footprint at
the expense of economic growth? An empirical analysis of 120 countries. J. Clean.
Prod. 346, 131207 https://doi.org/10.1016/j.jclepro.2022.131207.
Lin, B., Huang, C., 2023. Promoting variable renewable energy integration: the
moderating effect of digitalization. Appl. Energy 337, 120891.
Liu, Y., Yang, Y., Li, H., Zhong, K., 2022. Digital economy development, industrial
structure upgrading and green total factor productivity: empirical evidence from
China’s cities. Int. J. Environ. Res. Publ. Health 19 (4), 2414. https://doi.org/
10.3390/ijerph19042414.
Luo, S., Yimamu, N., Li, Y., Wu, H., Irfan, M.M., Hao, Y., 2022. Digitalization and
Sustainable Development: How Could Digital Economy Development Improve Green
Innovation in China? Wiley, pp. 8–23. https://doi.org/10.1002/bse.3223.
Ma, Q., Tariq, M., Mahmood, H., Khan, Z., 2022. The nexus between digital economy and
carbon dioxide emissions in China: the moderating role of investments in research
and development. Technol. Soc. 68, 101910 https://doi.org/10.1016/j.
techsoc.2022.101910.
Machado, J.A., Silva, J.S., 2019. Quantiles via moments. J. Econom. 213 (1), 145–173.
Maddala, G.S., Wu, S., 1999. A comparative study of unit root tests with panel data and a
new simple test. Oxf. Bull. Econ. Stat. 61, 631–652.
Messina, D., 2020. Contribuciones determinadas a nivel nacional del sector eléctrico en
América Latina y el Caribe: análisis de la transición hacia el uso sostenible de las
fuentes energéticas.
Mohsin, M., Kamran, H.W., Atif Nawaz, M., Sajjad Hussain, M., Dahri, A.S., 2021.
Assessing the impact of transition from nonrenewable to renewable energy
consumption on economic growth-environmental nexus from developing Asian
economies. J. Environ. Manag. 284, 111999 https://doi.org/10.1016/j.
jenvman.2021.111999.
Murshed, M., Apergis, N., Alam, M.S., Khan, U., Mahmud, S., 2022. The impacts of
renewable energy, financial inclusivity, globalization, economic growth, and
urbanization on carbon productivity: evidence from net moderation and mediation
effects of energy efficiency gains. Renew. Energy 196, 824–838.
Nagasawa, T., Pillay, C., Beier, G., Fritzsche, K., Pougel, F., Takama, T., Bobashev, I.,
2017. Accelerating Clean Energy through Industry 4.0 Manufacturing the Next
Revolution. A Report of the United Nations Industrial Development Organization.
Nwaiwu, F., 2021. Digitalisation and sustainable energy transitions in Africa: assessing
the impact of policy and regulatory environments on the energy sector in Nigeria and
South Africa. Energy, Sustainability and Society 11 (1), 1–16.
Ocal, O., Aslan, A., 2013. Renewable energy consumption–economic growth nexus in
Turkey. Renewable Sustainable Energy Rev. 28, 494–499. https://doi.org/10.1016/
j.rser.2013.08.036.
39
Journal of Cleaner Production 418 (2023) 138146
OECD, 2020. Latin American Economic Outlook 2020: Digital Transformation for
Building Back Better. OECD Publishing, Paris. https://doi.org/10.1787/e6e864fb-
en.
OLADE, 2020. Panorama energético de América Latina y el Caribe.
Pablo-Romero, M.P., Sánchez-Braza, A., Romero, M.G.P., 2022. Renewable energy in
Latin America. AIMS Energy 10 (4), 695–717.
Pesaran, M.H., Ullah, A., Yamagata, T., 2006. A Bias-Adjusted Test of Error Cross Section
Dependence.
Pesaran, M.H., 2007. A simple panel unit root test in the presence of cross-section
dependence. J. Appl. Econom. 22 (2), 265–312.
Pesaran, M.H., Yamagata, T., 2008. Testing slope homogeneity in large panels.
J. Econom. 142 (1), 50–93.
Ramzan, M., Raza, S.A., Usman, M., Sharma, G.D., Iqbal, H.A., 2022. Environmental cost
of non-renewable energy and economic progress: do ICT and financial development
mitigate some burden? J. Clean. Prod. 333, 130066.
Ren, S., Hao, Y., Xu, L., Wu, H., Ba, N., 2021. Digitalization and energy: how does
Internet development affect China’s energy consumption? Energy Econ. 98, 105220
https://doi.org/10.1016/j.eneco.2021.105220.
Royston, J.P., 1983. A Simple method for evaluating the shapiro–francia W′ test of non-
normality. J. Roy. Stat. Soc.: Series D (The Statistician) 32 (3), 297–300.
Royston, P., 1992. Comment on sg3. 4 and an improved D’Agostino test. Stata Technical
Bulletin 1 (3).
Santarius, T., Pohl, J., Lange, S., 2020. Digitalization and the decoupling debate: can ICT
help to reduce environmental impacts while the economy keeps growing?
Sustainability 12 (18), 7496. https://doi.org/10.3390/su12187496.
Santiago, R., Fuinhas, J.A., Marques, A.C., 2020. The impact of globalization and
economic freedom on economic growth: the case of the Latin America and Caribbean
countries. Econ. Change Restruct. 53, 61–85.
Sebri, M., Ben-Salha, O., 2014. On the causal dynamics between economic growth,
renewable energy consumption, CO2 emissions and trade openness: fresh evidence
from BRICS countries. Renewable Sustainable Energy Rev. 39, 14–23. https://doi.
org/10.1016/j.rser.2014.07.033.
Shahbaz, M., Wang, J., Dong, K., Zhao, J., 2022. The impact of digital economy on
energy transition across the globe: the mediating role of government governance.
Renewable Sustainable Energy Rev. 166, 112620 https://doi.org/10.1016/j.
rser.2022.112620.
Sharma, G.D., Shah, M.I., Shahzad, U., Jain, M., Chopra, R., 2021 a. Exploring the nexus
between agriculture and greenhouse gas emissions in BIMSTEC region: the role of
renewable energy and human capital as moderators. J. Environ. Manag. 297,
113316.
Sharma, G.D., Tiwari, A.K., Erkut, B., Mundi, H.S., 2021 b. Exploring the nexus between
non-renewable and renewable energy consumptions and economic development:
evidence from panel estimations. Renew. Sustain. Energy Rev. 146, 111152.
Singh, H.V., Bocca, R., Gomez, P., Dahlke, S., Bazilian, M., 2019. The energy transitions
index: an analytic framework for understanding the evolving global energy system.
Energy Strategy Rev. 26, 100382.
Smil, V., 2020. Energy transitions: fundamentals in six points. Papeles de Energia 8,
11–20.
Trinh, H.H., Sharma, G.D., Tiwari, A.K., Vo, D.T.H., 2022. Examining the heterogeneity
of financial development in the energy-environment nexus in the era of climate
change: novel evidence around the world. Energy Econ. 116, 106415.
Tugcu, C.T., Ozturk, I., Aslan, A., 2012. Renewable and non-renewable energy
consumption and economic growth relationship revisited: evidence from G7
countries. Energy Econ. 34 (6), 1942–1950. https://doi.org/10.1016/j.
eneco.2012.08.021.
Urban, F., 2014. Low Carbon Transitions for Developing Countries. Routledge.
Vanegas Cantarero, M.M., 2020. Of renewable energy, energy democracy, and
sustainable development: a roadmap to accelerate the energy transition in
developing countries. Energy Res. Social Sci. 70, 101716 https://doi.org/10.1016/j.
erss.2020.101716.
Višković, A., Franki, V., Bašić-Šiško, A., 2022. City-level transition to low-carbon
economy. Energies 15 (5), 1737. https://doi.org/10.3390/en15051737.
Wang, B., Wang, J., Dong, K., Dong, X., 2023. Is the digital economy conducive to the
development of renewable energy in Asia? Energy Pol. 173, 113381 https://doi.org/
10.1016/j.enpol.2022.113381.
Wang, J., Dong, K., Dong, X., Taghizadeh-Hesary, F., 2022a. Assessing the digital
economy and its carbon-mitigation effects: the case of China. Energy Econ. 113
https://doi.org/10.1016/j.eneco.2022.106198.
Wang, J., Wang, K., Dong, K., Shahbaz, M., 2022b. How Does the Digital Economy
Accelerate Global Energy Justice? Mechanism Discussion and Empirical Test.
Elsevier BV. https://doi.org/10.1016/j.eneco.2022.106315.
Wang, Q., Wang, L., 2020. Renewable Energy Consumption and Economic Growth in
OECD Countries: A Nonlinear Panel Data Analysis, vol. 207. Energy, Oxford),
118200. https://doi.org/10.1016/j.energy.2020.118200.
Wang, Q., Wang, L., Li, R., 2022c. Renewable energy and economic growth revisited: the
dual roles of resource dependence and anticorruption regulation. J. Clean. Prod.
337, 130514 https://doi.org/10.1016/j.jclepro.2022.130514.
Wirawan, H., Gultom, Y.M., 2021. The effects of renewable energy-based village grid
electrification on poverty reduction in remote areas: the case of Indonesia. Energy
for Sustainable Development 62, 186–194.
World Bank, 2022. Consolidating the Recovery: Seizing Green Growth Opportunities.
The World Bank.
Y.K. Hwang Journal of Cleaner Production 418 (2023) 138146

Xu, Q., Zhong, M., Li, X., 2022. How does digitalization affect energy? International
evidence. Energy Econ. 107, 105879 https://doi.org/10.1016/j.eneco.2022.105879.
Zhang, H., 2023. Pathways to carbon neutrality in major exporting countries: the
threshold effect of digital transition. Environ. Sci. Pollut. Res. Int. 1 https://doi.org/
10.1007/s11356-022-22592-x.
Zhang, W., Liu, X., Wang, D., Zhou, J., 2022. Digital economy and carbon emission
performance: evidence at China’s city level. Energy Pol. 165, 112927 https://doi.
org/10.1016/j.enpol.2022.112927.

40